Raven Pro 1.4 User`s Manual - Cornell Lab of Ornithology

Transcription

Raven Pro 1.4 User’s Manual
Revision 11
10 December 2010
This manual can be printed or used online. To access all online features, we recommend viewing with the latest version of Adobe [Acrobat]
Reader, available free at http://www.acrobat.com. For online viewing,
use the bookmarks in the navigation pane on the left to navigate to sections within the manual. Cross references in the text and index entries
are hypertext links: click the link to view the referenced section or page.
Copyright notice
Raven software, the Raven 1.4 User’s Manual, and example sounds Copyright ©2003 Cornell Lab of
Ornithology. All rights reserved. Reproduction of any part of this work in any form without permission is
prohibited.
The Raven software includes code licensed from RSA Security, Inc. Some portions licensed from IBM are
available at http://oss.software.ibm.com/icu4j/.
Trademarks
Microsoft and Windows are registered trademarks of Microsoft Corporation in the United States and/or
other countries. Mac and Mac OS are registered trademarks of Apple Computer, Inc. Linux is a registered
trademark of Linus Torvalds. PostScript is a registered trademark of Adobe Systems, Inc. Java is a
registered trademark of Sun Microsystems, Inc. Other brands are trademarks of their respective holders
and should be noted as such.
Mention of specific software or hardware products in this manual is for informational purposes only and
does not imply endorsement or recommendation of any product. Cornell University and the Cornell Lab
of Ornithology make no claims regarding the performance of these products.
Credits
Raven was developed with partial support from the US National Science Foundation (grant
DBI-9876714, Principal Investigators: Christopher W. Clark and Kurt M. Fristrup), with additional support
from the Cornell Lab of Ornithology.
Raven software was written by Harold Mills, Tim Krein, Dean Hawthorne, Scott Maher, Andrew Jackson,
Aisha Thorn, Dounan Hu, Laura Strickman, Christina Ahrens, Jason Rohrer, and Jason Adaska.
The Raven User’s Manual was written by Russell A. Charif, Amanda M. Waack, and Laura M. Strickman
with contributions by Tim Krein, Dean Hawthorne, Ann Warde, Dimitri Ponirakis, Wendy Alberg, and
Sarah Smith.
Raven artwork by Diane Tessaglia-Hymes.
The Raven development project is led by Tim Krein and is under the general direction of Christopher W.
Clark.
Citation
When citing use of Raven in scientific publications, please refer to the software by referring to the Raven
website: www.birds.cornell.edu/raven
When citing use of this manual, please refer to it as follows:
Charif, RA, AM Waack, and LM Strickman. 2010. Raven Pro 1.4 User’s Manual. Cornell Lab of
Ornithology, Ithaca, NY.
For more information about Raven, visit the Raven website: www.birds.cornell.edu/raven
Bioacoustics Research Program
Cornell Lab of Ornithology
159 Sapsucker Woods Rd.
Ithaca, NY 14850
USA
Raven 1.4 User’s Manual, revision 8
Table of Contents, Raven 1.4
Chapter 1 Getting Started . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1
About this chapter. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
The Raven window . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Menu bar . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Toolbars . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Side panel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Raven desktop . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1
1
1
2
Opening a sound file . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
Window preset and paging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Speed correction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
File format . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Opening sounds in different file formats . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
................................................................
Opening multi-channel files . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Opening files from Canary. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4
5
5
6
7
7
7
Understanding the Sound Window. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
Playing a sound . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
Making a selection. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Playing a selected part of a signal. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Filtered play. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Scrolling playback and position markers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
Looping and reverse playback . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
Multi-channel playback . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
Playback rate. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
Copying Part of a Sound . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
Copying a selection to a new sound window. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
Copy a selection to an existing sound. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
Saving All or Part of a Sound . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
Saving a signal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
Saving a selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
Saving Your Workspace . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
Saving a workspace . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
Opening a workspace . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
Printing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
Exporting images to files . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
Copying images to the clipboard . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
Recording a sound (acquiring input) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
Selecting an audio input device. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Create new recorder . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
The Recorder Window. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Starting and stopping the real-time signal display. . . . . . . . . . . . . . . . . . . . . . . . .
While recording is stopped… . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
More about recording . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
19
19
20
21
21
22
Chapter 2 The Raven Window . . . . . . . . . . . . . . . . . . . . . . . . . . . .23
i
About this chapter. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
The Menu bar. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
The File menu . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
The Edit menu . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
The View menu . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
The Window menu . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
The Tools menu . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
The Help menu . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
24
26
28
30
31
31
The Toolbars . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
The file toolbar. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
The edit toolbar . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
The view toolbar . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
The play toolbar. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
The spectrogram toolbar . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
32
33
34
36
37
Keyboard Shortcuts. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
The Side Panel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
Docking controls . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Vertical separator bar . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Show all/Hide all . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Side panel tabs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Information panel. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Mouse measurement field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
40
41
41
42
42
43
Changing the Appearance of the Raven Window . . . . . . . . . . . . . . . . . . . . . . 44
Selectable look and feel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
Selectable desktop background color . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
Selectable tooltip color . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
Chapter 3 Sound Windows:
Visibility, Views, Linkage, & Navigation . . . . . . . . . .45
Using Contextual Menus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
Basic Layout of a Sound Window . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
The active view . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Moving, resizing, and closing a sound window . . . . . . . . . . . . . . . . . . . . . . . . . . .
Scrollbars . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Axis units in views . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Major and minor grid lines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Position markers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Centering a position . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Positions control view appearance and behavior . . . . . . . . . . . . . . . . . . . . . . . . .
Position vs. position location . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Scale of a view . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Setting the scale of view axes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Changing view scales by zooming . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Zoom details . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Zooming in horizontally . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Zooming vertically . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Zoom to all. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Zoom to selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
ii
47
47
48
49
50
50
50
51
51
52
52
53
54
54
55
55
55
View Types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
The six main view types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Making a new view . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Showing data in a new view . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Selected view dictates toolbar content . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
The zoom window . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
56
57
59
60
60
Linking and unlinking views . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
Unlinked views . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Unlinking views through the contextual menu . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Restoring or creating new linkages through the linkage tab. . . . . . . . . . . . . . . . .
Creating a new linkage group . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Creating new linked views. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Other linkable properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Linkage between spectrogram and spectrogram slice views . . . . . . . . . . . . . . . .
62
63
64
65
65
65
66
Controlling how views are displayed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
Deleting a view . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
Changing the order in which views are displayed . . . . . . . . . . . . . . . . . . . . . . . . . 67
Hiding and showing window components . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
Changing the Appearance of a Sound Window . . . . . . . . . . . . . . . . . . . . . . . . 68
Spectrogram brightness and contrast . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Color schemes of sound windows. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Editing color schemes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Sound Window Presets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Multiple-line views within sound windows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Grouping views with multiple lines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Working with more than one sound. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Tile/Cascade windows. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Apply current window layout to all windows . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
68
69
71
71
74
75
77
78
80
Chapter 4 Signal Acquisition (Recording) . . . . . . . . . . . . . . . . . .81
Overview of signal acquisition. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
Recording modes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
Creating a recorder . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
Recording Input. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
Device . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Sample Rate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Channels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Sample Format . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
83
83
83
84
Recording Display. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
Buffer Size. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Update Rate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Latency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Window preset. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
84
85
85
86
File Format . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
File Format . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
Sample Size . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
File Size. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
File naming . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
Directory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
iii
Sound File(s) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
Start Time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
File name collisions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
Recording to a file sequence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
Using recorder presets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
The recorder window: real-time views during recording . . . . . . . . . . . . . . . 94
Creating real-time signal views . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Recording to memory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Recording to files. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Adjusting recording level . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Playing sound while recording. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Navigating and selecting data in a running recorder . . . . . . . . . . . . . . . . . . . . . . .
Reconfiguring an existing recorder . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Multiple recorder windows. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
95
95
96
96
97
98
98
99
Advanced options: acquiring from recordings at altered input speeds . . . . 99
Advanced options: sample rate conversion during recording . . . . . . . . . . . 101
Why decimation is useful. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
Decimation Factor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
Advanced options: Amplified recording . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
Advanced options: Adaptive filtering. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
Advanced options: scheduled recording . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
Retroactive Recording Offset . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
Advanced options: retroactive recording to files . . . . . . . . . . . . . . . . . . . . . . 104
Advanced options: exporting clip files while recording . . . . . . . . . . . . . . . . 104
Problems with recording . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
Unsupported audio format. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
Frame rate monitor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
Chapter 5 Spectrographic Analysis . . . . . . . . . . . . . . . . . . . . . . .109
About this chapter. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
About spectrographic views . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
Configuring spectrographic views. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
Window type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Window size . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Beta (Kaiser window only). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3 dB Bandwidth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Choosing the window size. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Time grid: Window Overlap and
Hop Size . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Frequency grid spacing and DFT size. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Clipping level . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Spectrum averaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Apply and Auto-apply . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Spectrogram presets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
113
116
117
117
118
120
122
123
127
129
129
Spectrogram views . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
Brightness and contrast. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
Time alignment of spectrogram data. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
Smoothed vs. unsmoothed display . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
iv
Spectrogram slice views. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
Significance of the spectrum values . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
Smoothed vs. unsmoothed display . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
Selection spectrum views . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
Significance of the spectrum values . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
Smoothed vs unsmoothed display . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
Changing parameters for an existing spectrographic view. . . . . . . . . . . . . . 140
Chapter 6 Selections:
Measurements, Annotations, & Editing . . . . . . . . . .141
About this Chapter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
Selection Basics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
Creating and modifying range selections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Point selections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Deactivating and clearing selections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Renumbering selections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
142
144
145
146
Editing a sound . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
Undoing changes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Filtering and amplifying sounds. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Defining your own filters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Adaptive Filtering. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Amplifying . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Batch Filtering and Amplifying. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
147
147
147
148
149
150
Selection Tables. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
The Selection Table . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Using selection tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Visibility of view and channel entries. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Sort order of view and channel entries. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Saving the selection table to a text file . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Retrieving selections from file . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Format of selection files . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Splitting a selection table. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
150
152
153
154
155
156
156
157
Working with selections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
Auto-advancing through selections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Viewing selections in the zoom Window . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Finding selections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Modifying selection bounds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Modifying selection channels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Selection tables in paged sound windows. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Saving the active selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Saving all selections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Highlighting selections. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Copying selected rows from the selection table . . . . . . . . . . . . . . . . . . . . . . . . .
Copying or moving selections between selection tables . . . . . . . . . . . . . . . . . . .
Clearing selections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Splitting a selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Generating multiple back-to-back selections. . . . . . . . . . . . . . . . . . . . . . . . . . . .
Creating duplicate selections at different times. . . . . . . . . . . . . . . . . . . . . . . . . .
Cloning selections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
157
158
158
158
160
160
160
161
161
161
162
162
163
165
166
166
v
Measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166
Choosing measurements to display . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Plottable Measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Measurements based on spectrogram values . . . . . . . . . . . . . . . . . . . . . . . . . .
Robust signal measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Measurements based on waveform values. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Selection-based measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Other measurements. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Measurement precision and format. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Using measurement presets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
166
167
168
170
174
175
176
177
177
Exporting samples from various views to text files . . . . . . . . . . . . . . . . . . . . 178
Export output and text file content. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
Annotations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
Creating, renaming, and deleting annotation columns . . . . . . . . . . . . . . . . . . . . 180
Entering annotations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
Selection Labels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
Pitch tracking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
Chapter 7 Large Datasets. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .187
About this Chapter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
Paging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
Configuring a new paged sound window. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Navigating through signals in paged sound windows . . . . . . . . . . . . . . . . . . . . .
Selections spanning pages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Reconfiguring paging for an existing window . . . . . . . . . . . . . . . . . . . . . . . . . . .
188
189
191
191
Batch Operations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
Background Tasks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
Progress Manager. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
Background task indicator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194
Opening file sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194
Using a list file . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
Selecting files in a sequence individually . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196
Opening multiple sound files in separate windows . . . . . . . . . . . . . . . . . . . 198
Correcting sounds acquired at altered speeds . . . . . . . . . . . . . . . . . . . . . . . . . 201
Correcting speed when opening file . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
Chapter 8 Multi-channel Sounds . . . . . . . . . . . . . . . . . . . . . . . . .205
Extensible multi-channel audio input device support . . . . . . . . . . . . . . . . . . . . . 205
Working with multi-channel files . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
Opening a multi-channel file . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Displaying and hiding channels. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Editing multi-channel sounds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Saving a subset of channels from an open signal. . . . . . . . . . . . . . . . . . . . . . . .
Playing multi-channel files. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
205
206
207
209
210
Exporting channels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
Saving selected channels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
vi
Batch channel export. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
Creating selections in multi-channel sounds . . . . . . . . . . . . . . . . . . . . . . . . . 212
Bearing analysis: Beamforming, Beamogram View, and Max Bearing Measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
Beamogram view. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Beamogram slice view . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Max bearing measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Configuring a beamogram. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Beamogram parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Hints for working with beamograms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
214
214
215
215
216
219
Chapter 9 Correlation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .221
Overview of Correlation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
Correlation types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
Calculating Correlation Functions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
Spectrogram correlations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
Waveform correlations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223
Using the correlation tool . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225
Band Pass Filter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Normalization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Biased/Unbiased . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Subset Normalization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
226
226
227
228
Spectrogram correlation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
Scale of spectrogram power values . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Demeaning of spectrogram values . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Spectrogram correlation parameters. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Effect of spectrogram clipping on correlations . . . . . . . . . . . . . . . . . . . . . . . . . .
..............................................................
232
233
234
235
237
Waveform correlations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237
Complex envelope. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
Batch correlation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 240
Input and output . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241
Running the correlator. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242
Correlator example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 246
Open sound file and selection table . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Save selections into a new folder . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Run batch spectrogram correlation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
The correlation table . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
246
247
248
249
Chapter 10 Detection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .253
The four run-modes of Raven detectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253
Interactive detection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255
Running a detector on an existing sound file . . . . . . . . . . . . . . . . . . . . . . . . . . . 255
General Detector Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 256
Displaying detector information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257
vii
Running an interactive detector in a paged sound . . . . . . . . . . . . . . . . . . . . . . . 259
Detection options. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 259
Running multiple interactive detectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 260
Full detection mode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261
Batch detection. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 262
Choosing files . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Choosing a detector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Configuring the detector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Saving the detection results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Configuring the view . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Running the batch detector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
263
264
264
264
265
265
Real-time detection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267
Running a detector within a recorder . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267
Example of running a detector within a recorder. . . . . . . . . . . . . . . . . . . . . . . . . 267
Saving detected selections. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273
The Band Limited Energy Detector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 274
Band Limited Energy Detector (Energy Detector) . . . . . . . . . . . . . . . . . . . . . . . . 274
Configuring the Band Limited Energy Detector . . . . . . . . . . . . . . . . . . . . . . . . . . 274
Band Limited Energy Detector Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 277
The Amplitude Detector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283
Configuring the Amplitude Detector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283
Chapter 11 Customizing Raven . . . . . . . . . . . . . . . . . . . . . . . . . . .289
Editing color schemes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 289
Choosing a color scheme element to edit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Choosing colors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Opacity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Color scheme presets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Auto-Apply, Apply, Reset, OK, and Close . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Creating customized colormaps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
293
293
294
295
296
296
The Preset Manager . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 297
About Raven preferences. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 299
Default presets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 301
Spectrogram presets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Recorder presets. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Measurement presets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Sound window presets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Color scheme presets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Selectable look and feel, desktop background color, tooltip color. . . . . . . . . . . .
301
302
302
303
303
304
Power Spectra Calculation Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Default workspace . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Default directories. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Default measurement precision . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
304
305
306
307
Time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 308
Frequency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 308
Amplitude . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 309
viii
Power . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 309
Overwrite behavior for recorders. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 309
The Memory Manager . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 309
About Raven memory allocation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 310
The Memory Manager window . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 311
Raven program and documentation updates. . . . . . . . . . . . . . . . . . . . . . . . . . 313
Automatic updates. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 313
Installing updates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 313
Troubleshooting Raven . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 314
Online resources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Reporting a bug. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Performance tips . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Contacting the Raven development and support team . . . . . . . . . . . . . . . . . . . .
..............................................................
Bug reporting. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Outgoing mail server . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Problems sending feedback . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
314
314
314
315
315
316
316
317
Appendix A Digital Representation of Sound . . . . . . . . . . . . . . . .319
About this appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 319
Digital sampling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 319
Sample rate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 320
Aliasing and the Nyquist frequency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 320
Sample size (amplitude resolution). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 323
Specifying sample sizes when acquiring and saving signals . . . . . . . . . . . . . . . 324
Storage requirements. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 325
Appendix B A Biologist’s Introduction to
Spectrum Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . .327
What sound is. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 327
Time domain and frequency domain representations of sound . . . . . . . . . 328
Spectral analysis of time-varying signals: spectrograms and STFT analysis .
331
Record length, bandwidth, and the time-frequency uncertainty principle 333
Making spectrograms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 334
Analysis resolution and the time-frequency uncertainty principle . . . . . . . . . . . .
Time grid spacing and window overlap . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Frequency grid spacing and DFT size. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Spectral smearing and sidelobes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Window functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
335
338
339
340
342
For further reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 344
ix
Appendix C Configuring Audio Input . . . . . . . . . . . . . . . . . . . . . .347
Windows 98, 2000, and XP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 347
Mac OS X. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 349
Appendix D Detector Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .351
Tradeoffs involved with using detectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 351
Results of detecting in less-than-perfect recordings . . . . . . . . . . . . . . . . . . . . . . 351
Receiver operating characteristic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 352
Changing a detector parameter. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 353
x
Chapter 1
Getting Started
About this chapter
This chapter introduces the major features of Raven using some of the
sound files supplied with the program as examples. Many details of how
these features work are deferred until later chapters. We recommend that
you read this chapter at the computer and try the examples as they’re
discussed. We assume you know how to browse for files on your
computer, and have installed Raven as discussed in the Raven README
file.
In this chapter you’ll learn how to:
•understand the basic layout of the Raven window
•open an existing sound file using Raven
•understand the sound window
•play all or part of a sound file
•copy part of a sound file
•save all or part of a sound file
•save your workspace
•record new signals
The Raven window
Launch Raven by double-clicking on its icon. (Depending on how you
installed Raven, there may be an icon for it on your desktop, in your Start
menu (Windows), in your dock (Mac OS), in your Applications menu
(Linux), or you may have to open the Raven folder to find it.) After
launching the Raven software, you will be greeted by the Raven window
(Figure 1.1) with its four main components.
Menu bar Raven’s menu bar appears just below the title bar of the Raven window on
Windows and Linux computers and at the top of the screen on Mac OS
computers.
Toolbars
Raven’s Toolbars contain controls for common Raven operations. When
you position the mouse pointer over a control in the toolbar, a “tooltip”
appears giving the name of the tool.
Side panel On the left side of the Raven window is the side panel. The side panel
contains tools that let you control various aspects of how signals are
displayed and played back.
1
Chapter 1: Getting Started
Raven desktop To the right, the main area of the screen is the Raven desktop. This area is
your workspace and will contain open sounds and other files.
Figure 1.1 The Raven window
Figure 1.1. The Raven window, showing the four main components
(menu bar, Toolbars side panel, and the desktop.)
For more information on the Raven window layout and functionality,
please visit Chapter 2, “The Raven Window”.
Opening a sound file
To open a sound file, choose File > Open Sound Files... , or type <Ctrl-O>
(Windows, Linux) or <Command-O> (Mac OS). You will then see Raven’s
Open Sound Files dialog box (Figure 1.2).
2
Figure 1.2 Open Sound Files dialog
Figure 1.2. Open Sound Files dialog box on a Windows computer (top)
and a Mac OS computer (bottom).
The Open Sound Files dialog box displays a scrolling list of the files and
directories in the current directory. The name of the current directory is
given at the top of the dialog box. If the current directory is not
“Examples”, find the Examples directory now in the Raven program
directory. (To move up in the file system hierarchy, click on the name of
3
the current directory and select another directory from the pull-down
menu that appears. To move down the hierarchy, double-click on the
name of a directory in the scrolling list of files and directories within the
current directory.)
WINDOWS
and LINUX
On Windows and Linux computers, the Audio File Format panel on the right side of the Open Sound Files
dialog box displays information about the currently
selected file.
MAC OS
On Mac OS, the file information is displayed on the lefthand side of the dialog box.
Select the file named “ChestnutSidedWarbler.aif” in the Examples
directory. Click the Open button, or double-click on the filename. The
Configure New Sound Window dialog box appears (Figure 1.3).
Figure 1.3. Configure New Sound Window dialog
Figure 1.3. The Configure New Sound Window dialog box
Window preset and The Configure New Sound Window dialog box allows you to choose a
paging window preset that controls the layout of a sound window (for more
information on sound window presets, see “Sound window presets” in
4
Chapter 11 (page 303)), and to control how much of the sound is loaded
into Raven’s working memory at one time (see “About Raven memory
allocation” in Chapter 11 (page 310) for more information on how Raven
uses memory.)
Speed correction The Speed tab of the dialog box gives you options to correct the recorded
speed of a stored sound. For more information on this, see “Correcting
speed when opening file” in Chapter 7 (page 201).
For now, click OK at the bottom of the dialog box to accept the default
settings. A sound window appears on the Raven desktop (Figure 1.4).
File format The Format tab of the dialog box lists the formatting information
associated with the chosen file. This information cannot be configured
from this dialog, but may serve as a useful reference when choosing other
settings.
Figure 1.4 Sound window
Figure 1.4. A sound window shown on the Raven desktop.
5
The title bar of the sound window shows a sequential number (starting at
1) that Raven assigns to each sound you open, and the name of the file.
When Raven is running, you can open any WAVE or AIFF
sound file by dragging its icon from an Explorer (Windows) or
Finder (Mac OS) window or from the desktop and “dropping”
the icon anywhere in the Raven desktop. If you hold down the
Ctrl key (Command key on a Mac) while dropping the file,
Raven will skip the Configuration dialog and open the sound
with the current settings.
To open a recently used sound file, choose File > Open
Recent Sound Files> fileName. By default, Raven displays
the four most recent files. To open a file from a folder that you
recently used within Raven, choose File > Open Recent
Folder > folderName.
Opening sounds in Raven also enables you to open sound files in .mp3 (with the exception of
different file formats variable bitrate mp3 files), .mp4, .aifc, and QuickTime movie soundtrack
(.mov) formats. In Raven Pro 1.4 build 33 and later, Raven opens these
files without any additional setup. In versions of Raven earlier than build
33, in order to access files in these formats you must have QuickTime (QT)
for Java installed on your computer. Raven can also open sound files in the
.flac format, which does not require the installation of QuickTime. Raven
does not include the ability to save sound files in these additional formats.
6
WINDOWS
You can check whether QT for Java is installed by looking at the Download QuickTime Installer item on the
Help menu. If QT for Java is already present, the menu
item will be disabled. If the menu item is enabled, click
on it and follow the instructions that appear in your web
browser to install QT for Java.
MAC OS
QT for Java is already installed on Mac OS computers.
Mac OS users can also open CD audio tracks with
Raven. Audio CD tracks will be displayed in the /Volumes folder as AIFC files that can be accessed from
Raven’s Open Files dialog.
Opening multi- Raven can open files that contain any number of channels. For informachannel files tion about how to open multi-channel files, see “Opening a multi-channel
file” in Chapter 8 (page 205).
Opening files from Raven Pro can open sound files saved in the Canary file format. Under
Canary Mac OS X, Raven Pro can open these files directly. In order to open Canary
files in Raven Fro on a computer running Windows, you must first convert
the files to MacBinary format on a Mac OS computer, then copy the files to
the Windows machine. You can convert Canary files to MacBinary by
using a free conversion program available at:
http://www.birds.cornell.edu/brp/software/canary-converters
Converting the files to MacBinary format is necessary
because Mac files have a two-part structure that is not properly preserved if the file is simply copied from a Mac to a Windows machine. The MacBinary format stores both parts of a
Mac file in a single file that can be transferred to a Windows
computer with no loss of information.
Note that while Raven Pro can open sound files saved by Canary in the
Canary file format and in the AIFF file format, Raven Pro cannot open files
saved by Canary in the MATLAB, SoundEdit, Text, or Binary formats.
Understanding the Sound Window
By default, when you first open a sound file, Raven shows you a sound
window that contains two views of the sound (see Figure 1.5). The waveform
(upper) view displays an oscillogram, or graph of the sound showing
amplitude versus time. The spectrogram (lower) view represents time on
the horizontal axis, frequency on the vertical axis, and relative power at
each time and frequency as a color (by default grayscale) value.
Spectrogram views are discussed further in “Spectrogram views” in
Chapter 5 (page 129).
7
Figure 1.5 Sound window
Figure 1.5. A sound window showing a waveform view (top) and a
spectrogram view (bottom).
In addition to waveforms and spectrograms, Raven can also display
spectrogram slice views of a signal, and selection spectrum views, which
show the average spectrum of a selected portion of a signal. Spectrogram
slice views and selection spectrum views are discussed in “View Types” in
Chapter 3 (page 56). Chapter 3, “Sound Windows: Visibility, Views,
Linkage, & Navigation” also explains how to configure Raven to display
combinations of views other than the default waveform and spectrogram
when a signal is first opened.
Playing a sound
The sound playback controls can be found in the right-hand end of the
play toolbar (Figure 1.6). Make sure the ChestnutSidedWarbler.aif sound
window is open, as directed in “Opening a sound file” on page 2. To play
the sound, click the Play button, or press <Ctrl-Shift-P> (Windows, Linux)
or <Command-Shift-P> (Mac OS).
Figure 1.6
Figure 1.6. The playback controls in the play toolbar.
8
As the sound plays, a vertical green line, the playback cursor, moves across
the waveform from left to right to show you what part of the signal you
are hearing. To stop playing at any time, click the Stop button. When the
selection finishes playing, or when you click Stop, the playback cursor
disappears.
You can change both the color and width of the playback cursor. The color can be changed in the color scheme editor. To
configure the width of the playback cursor, open the Raven
preferences file and change the associated value to your
desired width (in pixels).
raven.ui.soundWindow.playbackCursor.width=1
For more information on editing color schemes and the
Raven preferences file, see Chapter 11, “Customizing Raven”
Making a selection To choose a portion of the signal, click and drag your mouse from one
point in the sound to another a point, in either the waveform or
spectrogram view. Raven will mark your selection with a colored rectangle.
For more about making selections, see Chapter 6, “Selections:
Measurements, Annotations, & Editing”.
Playing a selected Once you’ve made a selection, you can choose to play only that portion of
part of a signal the sound by clicking the Play button. Clicking the Play button with an
active selection will only play the visible portion of the selection. If the
active selection is not currently visible, the Play button will play the entire
visible portion of a sound. To play the entire visible portion of a sound
with a selection active, use the Play Visible button.
For most views, Raven will play the visible portion of the
sound unless there is an active selection marked. For spectrogram slice and selection spectrum view, if there are no
selections, Raven will play the entire sound. This holds true
for reverse playback as well.
Filtered play When filtered play is turned on, Raven plays only the frequencies within
the bounds of the selection, rather than playing the entire available frequency band of the sound. This can be used to listen to only the higher
harmonics of a sound, for example, or to listen only to a low-frequency
animal call and not the high-frequency call recorded at the same time.
9
Figure 1.7
Figure 1.7. Playing a selection with filtered play on (left) and with filtered play off (right). Note that the play cursor indicates which frequency band is being played.
Scrolling playback The vertical and horizontal magenta lines in each view are the position
and position markers. Each view in a sound window has two position markers which
markers identify unique locations along the horizontal and vertical axes. The
vertical line on the left-hand side of each view is the time axis marker
(which marks one specific time value.) Position markers are important for
several types of Raven operations, and are discussed further in “Major and
minor grid lines” in Chapter 3 (page 50).
Using the mouse, grab the time axis position marker (the vertical magenta
line) in either the waveform or spectrogram view and move it slightly
away from the left edge of the window (Figure 1.8). Next, click the
Scrolling Play button to play the signal. In scrolling playback, the signal
view scrolls from right to left beneath the position marker, like tape
moving past the playback head of a tape recorder. The position marker
identifies the point in time that you are hearing.
Looping and To loop a sound window or selection to hear it many times, click on the
reverse playback Looping Play button. This will loop the sound continuously until you
press stop. It is also possible to play a sound backwards by clicking on the
Reverse Play button.
Multi-channel For files containing more than one channel of sound, Raven allows you to
playback choose which channels should be played back. You can specify these
settings in the Playback tab of the side panel or use the toggle button in the
play toolbar to only play channels that contain an active selection. For
more information on playing multi-channel sounds, see “Playing multichannel files” in Chapter 8 (page 210).
10
Figure 1.8 sound window, pos marker moved.
Figure 1.8. The BeardedSeal sound window, with the time axis position
marker moved away from the left edge of the window. The window’s
scroll thumb is also labeled.
During scrolling playback, you can drag the scroll thumb to
move the signal so that a particular point of interest is at the
time position marker. The sound stops playing while you drag
the scroll thumb and resumes immediately at the new time
position when you release the scroll thumb.
Playback rate You can speed up or slow down the playback rate of the signal. By default
the playback rate is set at 1.0, or the same as the recording rate. Slowing
down the rate makes the sound lower in pitch and slower; speeding it up
makes it higher in pitch and faster. Try it now: type a number greater than
1 in the rate box to speed up or a decimal number between 0 and 1 to slow
down, press <Enter>, and then play the sound again.
11
Copying Part of a Sound
You can copy data in the active selection using commands on the Edit
menu or standard keyboard equivalents. When you copy a selection, a
copy of the selected data is put in the clipboard. The Paste command
inserts the contents of the clipboard at the time of the active selection in
the active sound window. If the active selection is a range (rather than a
point), the clipboard contents replace the data in the selection. If there is
no active selection, the Paste command is unavailable. Data on the
clipboard can be pasted into the same sound window or into a different
one.
Copying a selection Create a new empty sound window by choosing File > New > Sound
to a new sound Window or by typing <Ctrl-N> (Windows, Linux) or <Command-N> (Mac
window OS). Next, make a selection in the ChestnutSidedWarbler.aif sound. Go to
Edit > Copy to copy the data in the selection to the clipboard. Next, click
on the new sound window to make it active and select Edit > Paste. You
should now see the selection you made in the ChestnutSidedWarbler.aif
sound appear in the new sound window (see Figure 1.9).
Figure 1.9 Copy a Selection
Figure 1.9. A selection from ChestnutSidedWarbler.aif copied into a
new sound window.
12
Copy a selection to With both sound windows open, make a new selection in the
an existing sound ChestnutSidedWarbler.aif signal. Copy this selection by choosing Edit >
Copy, click in the middle of the second sound window to create an active
selection, and paste the new selection there by choosing Edit > Paste
(Figure 1.10).
If the sample rate of the data on the clipboard is not the same
as that of the destination sound, Raven displays a warning
and gives the option to continue or cancel the paste
operation. If the sample rates do not match, the resulting
sound will probably not be what you intended.
Figure 1.10 Copying a Selection into a Sound
Figure 1.10. A second selection from ChestnutSidedWarbler.aif copied
and pasted into the middle of the sound in the bottom window.
Saving All or Part of a Sound
You can save the active signal in a sound file, either in WAVE format
(filename extension.wav) or in AIFF (Audio Interchange File Format,
filename extension .aif) format. WAVE files can be opened by most other
13
programs that work with audio data. AIFF files can be opened by most
Macintosh programs that work with audio data, and some programs on
other platforms.
To choose a file format (WAVE or AIFF), and a sample size for the file to be
saved, use the Files of Type drop-down menu in the Save As... dialog box
(Figure 1.11). Choice of sample size is discussed in Appendix A, “Digital
Representation of Sound”.
Files saved in WAVE or AIFF formats contain only the actual
audio data. Information that is specifically for use with Raven
(such as selection tables and layout parameters) is not saved
in these files.
Saving a signal To save a sound, choose File > Save “Sound N” or File > Save “Sound N”
As... . Choosing Save “Sound N” simply saves the sound under the same
filename in the same location as the last time the sound was saved. If the
sound has never been saved, Raven asks you to specify a location and
name for the file. Choosing Save “Sound N” As... allows you to specify a
new location and/or name for the file to be saved (Figure 1.11).
Figure 1.11 Save As... Dialog Box
Figure 1.11. This is the Save As... dialog box. After choosing a sound
file or a sound selection to save, you can select a location for the file
and specify a file name.
14
To save a sequence of files as a list file, choose File > Save As List File.
This saves a text file containing the names of the files and their order in the
file sequence. Saving a file sequence as a list file allows you to quickly
open the same sequence of files at a later time simply by opening the list
file. For more on working with list files, see “Opening file sequences” in
Saving a selection After making a selection, within the ChestnutSidedWarbler.aif signal for
example, choose File > Save Active Selection As... and you will see the
Save As... dialog box (Figure 1.11). After entering a name for your
selection and clicking Save, the active selection is saved into a new file by
itself which can be opened at any time.
Saving Your Workspace
When you save a workspace, all aspects of Raven’s state are saved. By
saving a workspace first, you can quit Raven and resume your work later
exactly where you left off, even if you have many sound windows open.
Saving a When you save a workspace, all information about Raven’s state is saved,
workspace including all open signals and the size and placement of their windows.
To save the workspace, choose File > Save Workspace As... . Raven
workspace files can be saved anywhere, and must have a filename
extension of .wsp.
Opening a To open a workspace file, choose File > Open Workspace... . If you have
workspace any sound windows open, Raven will warn you that they will be lost
when the workspace file opens, and ask if you want to proceed. If there are
signals open with unsaved changes, Raven gives you the opportunity to
save them before opening the workspace file. Once the saved workspace
opens, Raven is completely restored to its state at the time the workspace
was saved.
You can also open a Raven workspace by dragging the workspace file’s icon from the computer’s desktop, a Windows
Explorer window, or from a Mac Finder window, and dropping
the icon on the Raven desktop.
15
Earlier versions of Raven could not open workspace files if
any of the sound files they included had been moved. Now, if
Raven cannot find a sound file needed for a workspace that is
opening, the program will present a dialog box (Figure 1.12)
listing the missing file(s) and will allow you to find the file by
browsing through the system.
If you copy or move a workspace file to a different computer,
you must also copy or move the sound files and selection
table files (and their parent folders) that are referenced by the
workspace in order for Raven to open the workspace properly.
16
Figure 1.12 Workspace Resolver Dialog
Figure 1.12. Raven’s Workspace Resolver Dialog (top). This appears if
you attempt to open a workspace file with missing or moved sound files.
If you wish to find the missing file, choosing Browse... displays the
Select Dialog (bottom) which allows you to browse for and select the
sound file.
Printing
To print the active sound window, choose File > Print One Page of “Sound
N”..., then click OK in the print dialog that appears. This prints the visible
portion of a sound on one printed page. If the image is too large to be
displayed on a single page, Raven rescales the image to fit the page.
However, smaller images are not enlarged when printing. For smaller
images, you may need to adjust the size of the sound window to achieve
the desired printing result. For more flexibility with printing Raven
images, try exporting them to the clipboard (see below).
17
Alternatively, choosing File > Print Pages of “Sound N” will print the
entire sound starting at the current location with each printed page
displaying a length of sound equal to the visible section. Thus, choosing
Print Pages is equivalent to printing one page of the sound, advancing
forward by the length of the visible sound, printing another single page,
and repeating this procedure until the end of the sound is reached. When
printing a paged sound, Raven prints a full final page, which may result in
some overlap between the last two images. If the entire sound is stored in
memory, this overlap does not occur, and the final page is printed as a
partial page.
By default, Raven prints in portrait orientation. To change to landscape
orientation, choose File > Printer Page Setup, and choose Landscape
orientation in the Page Setup dialog. The currently selected print
orientation is indicated by the orientation of the page icon next to Printer
Page Setup in the File menu.
If you click on the printer Properties button in the Print dialog,
a dialog appears that contains another set of controls for
choosing page orientation. These controls may not correctly
display or allow you to change the actual page orientation. To
change the page orientation, always use File > Printer Page
Setup.
Exporting images to To save an image of all or part of the Raven window as a graphics file,
files choose File > Export Image Of. A submenu appears showing the graphics
objects that Raven can copy: the entire Raven window, the active sound
window, all views within the active sound window, or each individual
view within the active window. Choose whichever object you want to
export from the submenu. In the Export Image dialog box that appears,
choose a graphics file format. Any of the four graphics objects can be
saved in PNG, TIFF, JPEG, or BMP format. Views can also be saved in
Encapsulated Post Script (EPS) format, but image scaling does not work
for this format.
Copying images to To copy an image of all or part of the Raven window so that you can paste
the clipboard it into a document in another program, choose Edit > Copy Image Of. A
submenu appears showing the four graphics objects that Raven can copy:
the entire Raven window, the active sound window, all views within the
active sound window, or the active view of the active window. Choose
whichever object you want to copy from the submenu. The copied image
can be pasted into documents in any program that works with graphic
images.
Recording a sound (acquiring input)
Raven obtains its audio input from an audio source (e.g., a tape recorder,
CD player, or microphone) connected to a particular port on an audio
18
input device installed on your computer (e.g., microphone or line input
port of an internal sound card or USB sound input device).
Raven can acquire a signal directly to a file, to a sequence of files, or to
memory only (without saving to disk). While Raven is acquiring input you
can see multiple views— waveforms, spectrograms, spectrogram slices—
scroll by in real time.
This section covers acquiring a signal to memory. Recording to a file or file
sequence is discussed in Chapter 4, “Signal Acquisition (Recording)” and
also covers decimating the input signal (acquiring it at a lower effective
sample rate).
Selecting an audio You use controls supplied by the operating system to choose and
input device configure the audio input device that Raven will use. Appendix A,
“Digital Representation of Sound” discusses how to choose a particular
audio input device on each operating system. Before proceeding further,
you should refer to the appendix to ensure that your system is properly
configured.
Connect an audio source (e.g., tape recorder, CD player, or microphone) to
the appropriate port of the audio input device you selected.
Create new To ready Raven for acquiring a signal, click on the Record button on the
recorder file toolbar (Figure 1.13). You can also press <Ctrl-R> (Windows, Linux) or
<Command-R> (Mac OS), or choose File > New > Recorder... Doing so will
display the Configure New Recorder dialog box (Figure 1.14).
Figure 1.13 Record button
Figure 1.13. The Record button, located in the file toolbar of the Raven
desktop.
19
Figure 1.14. New Recorder dialog.
Figure 1.14. Raven’s Configure New Recorder dialog box.
Use fields in the Configure New Recorder box to acquire the signal to
memory or to one or more files, to choose which input device and device
configuration to use, and to specify how to display the sound while it’s
being acquired. For now, to give you a feel for Raven’s capabilities, we’ll
go over how to record sounds to memory (without saving them to disk)
and what you can do with the signal as it’s coming in. Chapter 4, “Signal
Acquisition (Recording)” explains in detail what each of the fields in this
dialog box do. For now, just leave all the fields as they are and click OK or
press <Enter>.
The Recorder When you click OK or press <Enter> in the Configure New Recorder
Window dialog box, a new recorder window appears on the Raven desktop (Figure
1.15). A recorder window looks and behaves like any other Raven sound
window except that it has additional controls displayed in the status bar at
the bottom of the window.
20
Figure 1.15. Recorder window (to memory).
Figure 1.15. A new Recorder window, configured for recording one
channel to memory.
Starting and Click the triangular green Record-to-Memory button (Figure 1.15) to start a
stopping the real- real-time scrolling waveform display in the recorder window. When you
time signal display start providing an audio signal (by starting playback of a tape or CD, or by
speaking into a microphone), you should see a waveform appear at the
right edge of the waveform window and scroll across to the left. The
Record-to-Memory button is replaced by a square Stop Recording button,
and the status field next to the button displays the message “Recording to
memory”.
When the waveform reaches the left edge, the oldest data are discarded to
make room for the newest data. Time counts up from the right side and
scrolls across the view. Clicking the Stop Recording button stops
recording. If you click the button to start recording again, Raven clears the
Recorder window before beginning to display the new signal.
By default, when recording to memory only, Raven records into a 30second sound buffer. You can specify a longer sound buffer when
configuring the recorder (see “Buffer Size” in Chapter 4 (page 84)).
While recording is When you stop a recording, the most recent part of the signal remains
stopped… displayed on the screen. You can do anything with this signal fragment
21
that you can do with a signal in any other sound window: save it, make
selections from it, copy from it, print it— whatever you like. Remember
that if you have been recording for a while you will only have the latest
part of the signal to work with (only what can be displayed), and not the
entire signal from the point at which you began recording.
More about Chapter 4, “Signal Acquisition (Recording)” covers the recording process
recording in more depth. Read that chapter to find out:
•how to select and configure your input device
•how to record to a file or a sequence of files
•how to incorporate date- and time-stamps into names of acquired files
•how to acquire signals at lower sample rates than those available from an
audio input device (signal decimation)
•other operations to perform while recording
22
Chapter 2
The Raven Window
About this chapter
In this chapter you’ll learn about the three main parts of the Raven
Window and how to adjust the appearance of the application. Topics
include:
•the Menu bar
•the Toolbars
•keyboard shortcuts
•the Side panel and mouse measurement field
•changing the appearance of the Raven Window
The Menu bar
The menu bar appears below the title bar on the top left of the Raven
window on Windows and Linux computers and at the top of the screen on
Mac OS computers (Figure 2.1)
Figure 2.1 The Menu bar
Figure 2.1. The Menu bar on a Windows computer.
Each entry on the menu bar contains specific commands and controls to
help you utilize all Raven’s tools. You can navigate through menus with
the mouse, or using keyboard mnemonics by pressing Alt + the underlined
letter for an item (i.e. Alt + F to open the file menu). Many menu items also
have direct keyboard shortcuts.The following is a brief discussion of the
menu bar entries that is intended to familiarize you with each menu item
and its contents.
23
Chapter 2: The Raven Window
The File menu
Figure 2.2 The File menu
Figure 2.2. The contents of the File menu.
‘New’ section in File menu
The first entry is in the ‘New’ section of the File menu, which was
discussed in Chapter 1, “Getting Started”. You can choose to create a new
sound window, new recorder window or a new selection table. These
these menu options also have keyboard shortcuts which are listed to the
far right of each menu item (<Ctrl-N> for New Sound Window in Windows,
for example.)
24
‘Open’ section in File menu
The next section contains the ‘Open’ options, some of which were
discussed in Chapter 1, “Getting Started”. You can choose to open a sound
from an existing file or CD, or choose a recently used file or another sound
from a recently used folder. You can also open a saved selection table or
workspace (see discussion in “The Selection Table” in Chapter 6 (page
150)), recent selection table, or another selection table from a recently used
folder. Again, some of these ‘Open’ options have keyboard shortcuts listed
to the far right if you prefer using those.
‘Close’ section in File menu
The ‘Close’ section contains commands used to close files and other items
that you were working with. You can choose to close an individual sound
file, close a selection table (more on these in “Selection Tables” in Chapter
6 (page 150)) or to simply close all the files open on the Raven desktop.
‘Save’ section in File menu
There are many options for saving your work, several of which were
mentioned in Chapter 1, “Getting Started”. In the Save section, you can
save sound files, selection tables, and workspaces that have already been
saved somewhere. To save items for the first time, or to save a file under a
new name, choose Save as... “Saving as...” is available for sound files,
selection tables, active selections, all selections, selected channels, and
workspaces.
‘Print’ section in File menu
In the ‘Print’ section, you can choose to print a sound file, print sections of
the file, or make changes to the page setup which will adjust the printed
output. Remember that the printer Properties button in the Print dialog
may not correctly display or change the page orientation; page settings
should only be adjusted through the File > Page Setup... menu item.
‘Export’ section in File menu
This section has options for exporting images (discussed in “Exporting
images to files” in Chapter 1 (page 18)) as well as exporting samples from
sound file views (more on this in “Export output and text file content” in
Chapter 6 (page 179)).
‘Exit’ in File menu
When you are finished using Raven, you can exit the application by
choosing File > Exit from the menu bar or by clicking the system close icon
in the window title bar.
25
The Edit menu
Figure 2.3 The Edit menu
Figure 2.3. The contents of the Edit menu.
Undo/Redo items in Edit menu
After performing an edit operation, you may wish to undo your change.
To revert back to the previous state, choose Undo. In some cases, you may
wish to reapply a change that was just revoked. To remake a change that
was undone, choose Redo. Note that the action you will be undoing/
redoing will also be listed in the menu (for example, Undo Filter or Redo
Paste).
‘Editing’ section in Edit menu
The main section of the Edit menu contains commands to Cut or Copy a
selection to the clipboard, Paste a selection from the clipboard, or Delete a
selection. These editing options all deal with sound samples, and more
information about these processes can be found in “Editing a sound” in
Filter/Amplify items in Edit menu
Also in the Edit menu, you can choose to Filter and/or Amplify parts of
signals. More details on these processes can be found in “Filtering and
amplifying sounds” in Chapter 6 (page 147).
Copy Image Of
Similar to the Export Image Of... function in the File menu, you can choose
to copy image information to the clipboard from within the Edit menu.
Selecting Copy Image Of allows you to copy an image of the entire Raven
window, a particular sound window, or certain views within a sound
window to the clipboard for use in other applications.
Select All
26
Choosing this menu item selects all information displayed in the Raven
sound window. In cases where you need to make changes or edit all the
information displayed, this bypasses the time and effort required to
individually select everything. Also, if the sound window is zoomed to a
subset of the entire sound, only the visible subset of the current view will
be selected.
Preferences
This item opens the Raven preferences file for editing. The preferences file
stores many of Raven’s presets and default settings. This file is also saved
in the Raven program directory as RavenPreferences.txt. For more
information on editing the preferences file, see “About Raven preferences”
in Chapter 11 (page 299).
27
The View menu
Figure 2.4
Figure 2.4. The contents of the View menu.
Basic commands
From this menu, you can select which toolbars are visible and you can lock
their positions. You can create a new view or work with window presets
(see more on these topics in Chapter 3, “Sound Windows: Visibility, Views,
28
Linkage, & Navigation” and Chapter 11, “Customizing Raven”). You can
choose to apply the view settings of a sound window to all other open
windows and you can also adjust configuration settings of a recorder.
‘View’ section in View menu
This section contains commands that allow you to adjust views by hiding
them, deleting them, or moving them up and down within a sound
window. More information regarding views is discussed in Chapter 3,
“Sound Windows: Visibility, Views, Linkage, & Navigation”.
Unlinking
You can choose to unlink views within a sound window by selecting this
submenu. Details on how this works and when you might want to do this
can be found in “Linking and unlinking views” in Chapter 3 (page 62).
Configure view
Views can be adjusted in several ways. The color scheme can be changed,
and the display axes of the view can be adjusted or hidden. You can also
choose which, if any, grids to display in each view. For more information
on how to adjust views, see Chapter 3, “Sound Windows: Visibility,
Views, Linkage, & Navigation”. This section may also have more menu
items available depending on the active view (spectrogram views, for
example, will have entries allowing you to adjust spectrogram parameters
and smoothing.)
Detectors
You can select and run interactive detectors by choosing this item.
Measurements, annotations, and labels
You can choose to make measurements and annotations of displayed
information in the selection table, choose which selections to display, and
how these selections are labeled. More information on this can be found in
Chapter 6, “Selections: Measurements, Annotations, & Editing”.
Selections
After making selections, there are many things you can do with them,
including copying, clearing, and pasting them. Some of these commands
can be found in the Active Selection submenu. This section also includes
ways to locate or highlight a particular selection in the selection table.
Basic selection copying was discussed in “Copying Part of a Sound” in
Chapter 1 (page 12) and you can find more details on selections and
various ways to work with them in Chapter 6, “Selections: Measurements,
Annotations, & Editing”.
29
The Window menu
Figure 2.5 Window menu
Figure 2.5. Contents of the Window menu.
Management
This section of the Window menu contains advanced commands to control
aspects of Raven’s performance. You can find more information about
these options in Chapter 11, “Customizing Raven”. Also listed in this
section is the option to open a zoom window for viewing selections. More
information about the zoom window can be found in “The zoom
window” in Chapter 3 (page 60).
Window organization
Tile and Cascade Windows are two ways to organize sound windows on
the Raven desktop. Details about these commands can be found in “Tile/
Cascade windows” in Chapter 3 (page 78).
Appearance
The appearance of the Raven application and desktop area can be
adjusted. For more instructions on personalizing Raven’s appearance, see
“Selectable look and feel, desktop background color, tooltip color” in
Window Selection
If you have multiple sound windows open on the Raven desktop, they will
be listed in the Window menu. Choosing one of them from the bottom of
the Window menu will activate that window and bring it to the front. This
makes it easier to find exactly the sound window you need.
30
The Tools menu
Figure 2.6
Figure 2.6. Contents of the Tools menu.
The items in the Tools menu are advanced ways to work with data in
sound files. For information on using the correlation tool, please refer to
Chapter 9, “Correlation”. For help with making detections, see Chapter 10,
“Detection”. And to perform batch operations on many sounds at once,
see “Batch Operations” in Chapter 7 (page 192).
The Help menu
Figure 2.7 Help menu
Figure 2.7. Contents of the Help menu.
Documentation
This section of the Help menu contains links to documentation for the
Raven software. You can access the entire Raven User’s Manual by
selecting the first option or you can choose to view the “What’s New”
document. You can visit the Raven web site, view frequently asked
31
questions, or read the Raven license agreement. You can also access the
Raven Help Forum which is a great tool for users.
Feedback
By selecting the Email Feedback option, you can submit comments, bug
reports, support requests, and other feedback to the Raven software team
via email. For more information on contacting the Raven software team
and how to use the Email Feedback form, please see “Contacting the
Raven development and support team” in Chapter 11 (page 315).
Updates
Periodically, the Raven software team releases updates and changes to the
Raven application. You can check for and read about these updates in this
section of the Help menu. You can also download the QuickTime installer
(more on this process can be found in “Opening sounds in different file
formats” in Chapter 1 (page 6)).
About
Selecting About Raven... allows you to view information regarding
copyright information, thanks, and credits for the application.
The Toolbars
There are five main toolbars in Raven. You can control whether a toolbar is
displayed or hidden by selecting View > Toolbars > <Toolbar Name>. If
the toolbar has a checked box next to its name, it is already visible. If the
box is empty, the toolbar is hidden. In the same menu (View > Toolbars)
you can also choose to lock or unlock the toolbars. If the icon is a locked
padlock, the toolbars are locked and you will be unable to move or
otherwise adjust their positions. Unlocking the toolbars (denoted by an
unlocked padlock) allows you to move the toolbars around if you so
desire.
The file toolbar The file toolbar is located underneath the menu bar, across the top of the
Raven window (see Figure 2.8). It contains buttons allowing you to create
new items, open existing items, save items, and print (in addition to
adjusting the page setup for printing purposes). Since this section is
intended only to introduce the contents of the toolbar, specific information
about the commands and their functionality can be found throughout this
manual.
32
Figure 2.8 file toolbar
Figure 2.8. The file toolbar.
New Recorder, New Sound Window, New Selection Table
The first three buttons in the file toolbar create new items (recorders,
sound windows, or selection tables). You can also create these items by
accessing the File menu (“‘New’ section in File menu” on page 24).
Open Sound Files, Open CD, Open Selection Table, Open Workspace
The next set of four buttons allows you to open various items (sounds
from files or CD, selection tables, or saved workspaces). You can also open
these items from the File menu (“‘Open’ section in File menu” on page 25).
Save Sound, Save Sound File As..., Save Selection Table, Save Selection
Table, Save Active Selection As..., Save All Selections As..., Save Selected
Channels As..., Save Workspace
There are 7 buttons with various Save functionality in the toolbar. You can
save certain items through the File menu as well (“‘Save’ section in File
menu” on page 25).
Print Sound, Page Setup
If you’d like to print a sound, or adjust the page settings before printing,
the last two buttons provide an easy way to do so. Of course, you can
always go through the File menu to access these options as well (“‘Print’
section in File menu” on page 25).
The edit toolbar The edit toolbar is typically located underneath the file toolbar (although,
if you unlock the toolbars and move them around, the location could be
different). This toolbar contains buttons related to editing, filtering, and
amplifying sounds and selections. More information on these topics can
be found in Chapter 6, “Selections: Measurements, Annotations, &
Editing”.
33
Figure 2.9
Figure 2.9. The edit toolbar.
Undo, Redo, Cut, Copy, Paste, Delete, Select All
These buttons contain basic editing commands that can also be applied by
selecting the Edit menu (“Undo/Redo items in Edit menu” on page 26 and
“‘Editing’ section in Edit menu” on page 26).
Raven Preferences
This button provides a link to open and edit the Raven preferences file,
which stores many of Raven’s presets and default settings. This file is also
saved in the Raven program directory as RavenPreferences.txt. For more
information on editing the preferences file, see “About Raven preferences”
Filter Around Active Selection, Filter Out Active Selection, Filter Active
Selection With..., Filter Around All, Filter Out All, Filter All With..., Amplify...
To quickly perform filtering or amplification, you can use these buttons.
Alternately, you can perform the same tasks by accessing the Edit > Band
Filter menu or Edit > Amplify item (“Filter/Amplify items in Edit menu” on
page 26) if you prefer.
The view toolbar This toolbar contains buttons that will create new views as well as adjust
the way sound windows and views will appear. For more information on
these topics, see Chapter 3, “Sound Windows: Visibility, Views, Linkage,
& Navigation”.
34
Figure 2.10 the view toolbar
Figure 2.10. The view toolbar.
New Waveform View, New Spectrogram View, New Spectrogram Slice
View, New Selection Spectrum View, New Beamogram View, New
Beamogram Slice View, New Color Bar View
The first seven items on the view toolbar are the New Views buttons and
they are only accessible when a sound window is open. These buttons
create new waveform, spectrogram, spectrogram slice, selection spectrum,
and color bar views, respectively. When a multi-channel signal is open,
the beamogram and beamogram slice views also becomes available. You
can also create new views by accessing the View > New submenu (“‘View’
section in View menu” on page 29). For more information on what these
views are and how to use them, see “View Types” in Chapter 3 (page 56).
Create Selection Mode, Activate Selection Mode, Grab and Scroll Mode
These buttons allow you to choose between the three selection modes:
Create, Activate, or Grab And Scroll. Create Selection Mode is the default
mode which displays a crosshair and allows you to create selections in the
sound window. Activate Selection Mode allows you to easily activate a
selection in the current selection table by clicking on it in the sound view,
but does not allow you to create or modify selections. Grab And Scroll
Mode allows you to move through a sound by clicking and dragging a
particular point in the sound window. Using Grab And Scroll Mode while
35
holding down the shift key allows you to scroll horizontally while locking
the vertical axis.
Clear Selection, Clear All Selections, Zoom to Selection, Zoom to All, Zoom
In X, Zoom Out X, Zoom In Y, Zoom Out Y
If you want to quickly clear a single selection, or clear all selections in a
sound window, you can use the first two buttons on the left. You can also
access these commands at the bottom of the View menu, or by using the
selection context menu. If you’d like to adjust the appearance of a
particular view, you can use the zoom buttons in the toolbar. The same
functionality can be achieved by using the zoom buttons on the sound
window itself (“Changing view scales by zooming” in Chapter 3 (page
53)).
Progress Manager, Zoom Window
To use Raven’s progress manager to view the progress of background
tasks, use the Progress Manager button. For more information, see
“Progress Manager” in Chapter 7 (page 192). The Zoom Window button
can be used to examine a selection; see “The zoom window” in Chapter 3
(page 60).
Tile Windows, Tile Windows Horizontally, Tile Windows Vertically, Cascade
Windows
If you want to change how the sound windows are displayed on the
desktop, you can choose to tile or cascade them using the four buttons on
the far right. These commands can also be found listed in the Window
menu (“Tile/Cascade windows” in Chapter 3 (page 78)).
The play toolbar The play toolbar contains the playback controls. The use of these controls
was discussed in “Playing a sound” in Chapter 1 (page 8).
Figure 2.11
Figure 2.11. The play toolbar.
Play, Play Visible, Scrolling Play, Looping Play, Reverse Play, Stop,
Filtered Play
These buttons control the playback of a sound. For more information on
using these controls, see “Playing a sound” in Chapter 1 (page 8).
36
Only Play Channels with Active Selection
When playing multi-channel files, you can opt to only play channels that
contain the active selection. For more information on playing multichannel files, see “Playing multi-channel files” in Chapter 8 (page 210).
Playback Rate
This controls the speed at which a sound is played. You can speed up or
slow down the playback rate of the signal. By default the playback rate is
set at 1.0, or the same as the recording rate. Slowing down the rate makes
the sound lower in pitch and slower; speeding it up makes it higher in
pitch and faster.
The spectrogram The spectrogram toolbar gives you access to making quick adjustments to
toolbar the appearance of spectrogram displays. For more detailed information on
spectrograms, see Chapter 5, “Spectrographic Analysis”.
Figure 2.12 the spectrogram toolbar
Figure 2.12. The spectrogram toolbar.
Color Map: Changing and Reversing
To change the color map of your spectrogram, simply use the Change
Color Map drop-box. Choices include Grayscale (standard), Cool, Hot,
Standard Gamma II, Copper, and Bone. You can also choose to reverse the
color map. For more on spectrogram color maps, see “Color schemes of
sound windows” in Chapter 3 (page 69).
Brightness: Adjusting and Setting
To adjust the brightness of the spectrogram window, you can use the
adjustable slider (sliding the bar to the right increases brightness while
sliding it to the left decreases the brightness) or you can set the brightness
to an exact level by entering a number (1 to 100) into the Set Brightness
text box. For more information on this, please see “Spectrogram brightness
and contrast” in Chapter 3 (page 68).
Contrast: Adjusting and Setting
To adjust the contrast of the spectrogram, you can use the adjustable slider
(sliding the bar to the right increases the contrast while sliding it to the left
decreases the contrast) or you can set the contrast to an exact level by
entering a number (1 to 100) into the Set Contrast text box. For more
37
information on this, please see “Spectrogram brightness and contrast” in
Spectrogram Window Size: Adjusting and Setting
The window size parameter controls the length of each data record that is
analyzed to create each of the individual spectra that together constitute
the spectrogram. The default unit for this measurement is number of
samples. For more information on window size, please see “Window size”
Keyboard Shortcuts
Many of the functions listed in the menus and toolbars also have keyboard
shortcuts associated with them. A list of these functions, and their
corresponding shortcuts in Windows is provided below. The shortcuts for
Mac are the same as those for Windows, but use the Command key in
place of Ctrl.
Command
38
Windows Shortcut
New Sound Window
Ctrl + N
New Recorder
Ctrl + R
New Selection Table
Ctrl + T
New Selection Table (with Measurements)
Ctrl + Shift + T
Open Sound Files
Ctrl + O
Open Selection Table
Ctrl + Shift + O
Save Sound N
Ctrl + S
Save As List File...
Ctrl + Shift + F
Save Selection Table
Ctrl + Shift + S
Save Active Selection As...
Ctrl + Shift + V
Save All Selections in Current Table As...
Ctrl + Shift + A
Save Selected Channels As...
Ctrl + Shift + H
Print Pages of “Sound N”
Ctrl + P
Export “Sound N” “View M” Samples...
Ctrl + Shift + X
Undo
Ctrl + Z
Redo
Ctrl + Shift + Z
Cut
Ctrl + X
Command
Windows Shortcut
Copy
Ctrl + C
Paste
Ctrl + V
Delete
Ctrl + Delete
Select All
Ctrl + A
New Waveform View
Ctrl + W
New Spectrogram View
Ctrl + G
New Spectrogram Slice View
Ctrl + L
New Selection Spectrum View
Ctrl + M
New Beamogram View
Ctrl + B
New Color Bar View
Ctrl + C
New Similar Spectrogram View
Ctrl + Shift + G
New Similar Spectrogram Slice View
Ctrl + Shift + L
New Similar Selection Spectrum View
Ctrl + Shift + M
Play
Ctrl +Shift + P
Play Visible
Ctrl + Shift + Y
Reverse Play
Ctrl + Shift +R
Copy Selected Cells
Ctrl + Shift + C
Clear Selected Cells
Ctrl + Shift + Backspace
Activate Previous Selection
Ctrl +Up
Activate Next Selection
Ctrl + Down
Move Selection Control Point
Ctrl + Arrow
Switch Between Control Points
Tab
Open Sound
Drag icon into Raven Window
Open Sound (Skip Configure Dialog)
Ctrl + Drag icon into Raven
Window
Move Horizontal Position Indicator
Ctrl + Click
Scroll sound and position marker
Ctrl +Shift + Click
39
The Side Panel
The left side of the Raven window holds the side panel, which contains six
tabs: Layout, Linkage, Selection, Playback, Detection, and Information. By
default, the side panel opens to the Layout tab; see Figure 2.13.
Figure 2.13 The Side Panel
Figure 2.13. A general view of the side panel, with basic layout
features outlined in red.
Docking controls Clicking on these arrows controls the visibility of the side panel. If you
choose the arrow pointing toward the left (at the top for Windows and
Linux users, or at the bottom for Mac OS) the side panel will be docked or
hidden, and you will be unable to view its contents. If you select the arrow
pointing toward the right, the side panel will become undocked and will
40
become visible again. When the side panel is docked or undocked, any
maximized sound windows will be resized to occupy the entire Raven
desktop. Similar docking control arrows appear throughout the
application in cases where panels can be docked to provide more viewing
room for other information.
Vertical separator The gray shaded bar to the right of the side panel can be moved to the left
bar and right to control the width of the side panel itself. To resize the side
panel, simply click your mouse over the control bar and, while holding the
mouse button, drag the mouse to the right or left.
Show all/Hide all The side panel contains different sections of information. Each section in
the panel contains a list of items. When a checkbox next to an item is
checked, that item will be shown in the active sound window on the
Raven desktop. For example, if you open ChestnutSidedWarbler.aif, the
items listed under the View section in the side panel should be ‘Waveform
1’ and ‘Spectrogram 1’, with both checkboxes checked. This means that
both the waveform and spectrogram views are shown in the
ChestnutSidedWarbler.aif sound window. For a brief review of what the
sound window looks like, please refer to “Understanding the Sound
Window” in Chapter 1 (page 7).
Sometimes, you may want to hide all the items listed in a side panel
section. To do this, select the Hide all button (the three white, ‘empty’
boxes). This will uncheck the checkboxes and will hide these items from
being seen in the sound window. To show all the items listed in a side
panel section, select the Show all button (the three black boxes) and this
will make the items visible in the sound window. See Figure 2.14. for more
information on hiding and showing all components.
41
Figure 2.14.. Layout tab
Figure 2.14. The Layout tab for a sound window that contains two
views. The Component Visibility pane has been hidden by dragging the
separator bar above it down.
Side panel tabs At the bottom of the side panel are five tabs which dictate what
information is displayed in the side panel. Clicking on one of the tabs will
change the contents of the side panel to show information pertinent to the
tab’s heading. Details regarding these tabs and how they can be used will
be discussed at later points throughout this manual.
Information panel The sixth tab holds the Information Panel, which displays information
about the sound window that is currently open. It will tell you the sound’s
file type, number of channels, sample rate, encoding, and length. The
sample rate is adjusted to compensate for any speed changes; so a sound
recorded at half-speed at 44 kHz will be listed as having a sample rate of
88 kHz.
When a window contains more than one sound (for example, a correlation
window), the Information Panel displays information about the first
sound in the list.
42
Mouse At the bottom of the Raven window is an area called the mouse
measurement field measurement field which displays information from whatever sound
view your mouse pointer is currently positioned over. See Figure 2.15 For
a file sequence, the name of the file at the pointer location is also
displayed.
Figure 2.15 Mouse measurement field
Figure 2.15. The Raven window with the mouse measurement field
labeled in red.
Depending on what type of view your mouse pointer is positioned over,
different information will be displayed here.
•As you move the mouse pointer over a waveform view, the time of the
pointer’s location and the amplitude of the waveform at that time are
shown in the mouse measurement field.
•When moving the mouse pointer over a spectrogram view, the time and
frequency of the pointer’s location, and the relative power at that time
and frequency are shown in the mouse measurement field.
43
•As you move the mouse pointer across a spectrogram slice view, the
mouse measurement field at the bottom of the Raven window displays
the frequency at the mouse pointer location, and the relative intensity at
that frequency, for the time slice shown.
Often, the time scale is so compressed that the number of sample points
represented exceeds the number of pixels in the window. In this case, each
pixel stands for multiple sample points, and the measurement display
shows the minimum and maximum values of the samples represented at
the time of the mouse position (as in Figure 2.15).
Changing the Appearance of the Raven Window
Selectable look and The color and texture of the Raven window (the “look and feel”) can be
feel selected so that it mimics the appearance of several standard application
types. You can also choose to retain Raven’s unique style and appearance.
Choose Window > Look and Feel and select Metal, Motif, or Windows.
Metal is Raven 1.4’s standard appearance.
Selectable desktop Choose Window > Background Color... to open the Background Color
background color Editor panel and set the color of the main Raven screen area (the desktop)
through Swatches, HSB, or RGB color definitions. More about customizing
Raven can be found in Chapter 11, “Customizing Raven”.
Selectable tooltip Choose Window > Tooltip Color... to open the Tooltip Color Editor panel
color and set the color of Raven’s tooltips, the helpful signs that pop up when
you hover your mouse cursor over a button or box. Note that tooltip color
can only be changed when the Metal Look and Feel is in use.
44
Chapter 3
Sound Windows:
Visibility, Views, Linkage, & Navigation
About this chapter
This chapter will explain how to work with and get around Raven in more
detail. Before reading this chapter, you should have a solid understanding
of the major components and basic layout that make up the Raven
window. If you need to review this information, please see Chapter 1,
“Getting Started” and Chapter 2, “The Raven Window”. Topics discussed
in this chapter are:
•using contextual menus
•learning the basic layout of a sound window
•understanding the five main view types
•linking and unlinking of views
•controlling how views are displayed
•changing the appearance of a Sound window
•changing the appearance of the Raven window
Using Contextual Menus
As you learn more about navigating around the Raven window, you
might realize that it would be convenient to have a list of contextuallyoriented commands readily available. Well, luckily for our users, there is
such a list. A context menu is a menu of useful commands that apply to
wherever your mouse pointer is positioned at that time. To activate a
context menu, simply right-click using your mouse (or Control+click on a
Mac) and the menu will appear.
45
Chapter 3: Visibility, Views, & Navigation
Figure 3.1 Context menu example
Figure 3.1. An example of a context menu when the mouse pointer is
over a waveform view. The menu gives you many command options
relevant to the mouse location and is accessible by a simple right-click
of the mouse (or Control+click on a Mac).
Basic Layout of a Sound Window
This section contains information regarding the general makeup of sound
windows. To review how to open a sound window, refer to “Opening a
sound file” in Chapter 1 (page 2).
46
Figure 3.2 Sound Window
Figure 3.2. A basic sound window containing two views (a waveform
view, top, and a spectrogram view, bottom) with the view axes (both
vertical and horizontal) marked in red. Each axis is marked with a unit
and number values.
The active view Most commands, buttons and scroll bars that affect views apply only to
the active view. Only one view in a sound window can be active at any
given time. The active view is identified by a colored vertical view selection
button at its left edge. The name of the active view is also highlighted on
the side panel’s Layout tab. In Figure 3.2, the waveform is the active view.
To activate a view, click on its view selection button or in its axis areas. You
can also activate a view by clicking on its name to select it in the side
panel’s Layout tab.
Although you can activate a view by clicking in the data
region (i.e., above and to the right of the axes), doing so will
create a selection. Basic information about selections can be
found in “Making a selection” in Chapter 1 (page 9), while
more advanced selection information is discussed in Chapter
6, “Selections: Measurements, Annotations, & Editing”.
Moving, resizing, The sound window’s title bar has controls for resizing and closing the
and closing a sound window, and is the means by which you move a sound window. See
window Figure 3.3.
47
Figure 3.3 Moving and Resizing a Sound Window
Figure 3.3. The title bar, with controls to resize and close a sound
window. The title bar is also the anchor for moving the sound window.
To move the window around, simply click and drag on the title bar.
To minimize the window (to reduce the whole window to a short title bar
at the bottom of the Raven window) click on the Minimize icon in the title
bar. Clicking on a minimized window’s title bar expands it again.
Clicking on the maximize icon makes the window fill the entire Raven
desktop. When the window is maximized, you’ll notice the Maximize
button changes into a Restore button. Clicking this Restore icon restores
the sound window to its previous size and position.
Clicking on the Close icon closes the window; if you close it, you’ll have to
reopen the sound file again by choosing File > Open Sound Files..., by
typing <Ctrl-O> (Windows, Linux) or <Command-O> (Mac OS), or by
using the Recent Files section of the File menu.
You can also resize the window by clicking and dragging on an edge or
any corner of the window. On Mac OS, you must drag from the bottom
right corner to resize the window.
Scrollbars The horizontal and vertical scrollbars in a Raven sound window always
refer to the active view. The length of the horizontal scrollbar in a
waveform or spectrogram view corresponds to the total duration of the
sound that is in Raven’s working memory.1 The length of a scrollbar’s
scroll thumb (Figure 3.4) relative to the length of the entire scrollbar,
indicates what proportion of the corresponding axis is visible in the view
pane.
1. If you opened the entire sound at once (the default), the duration of the sound
in memory is the duration of the entire sound file or file sequence. If you
opened the sound in a paged window, the duration in memory is the length of
one page. See “Configuring a new paged sound window” in Chapter 7 (page
188) for more on paged sound windows.
48
Figure 3.4 Scrollbars
Figure 3.4. A sound window with its scrollbars and scroll thumbs
labeled in red.
As touched on in “Filtered play” in Chapter 1 (page 9) the location of the
scroll thumb within the scrollbar indicates the view’s position relative to
the data. When the horizontal scroll thumb of a waveform or spectrogram
is at the left edge of the scrollbar, the start of the data is aligned with the
position marker (Figure 3.4).
Raven displays a gray background for areas in each view
pane that are beyond the limits of the data, for example
before the beginning or after the end of a signal in the time
dimension.
Axis units in views The units used on the axes are indicated in the lower left corner of each
view. In the waveform, the units are seconds (S) for the horizontal time
axis, and kilounits (kU) for the vertical amplitude axis. In the spectrogram,
49
the units are seconds (S) for the horizontal time axis, and kilohertz (kHz)
for the vertical frequency axis.
The “units” displayed on the vertical axis of a waveform view
are the actual sample values in the signal, which are proportional to the sound pressure at the microphone when the
sound was recorded.
Major and minor You can choose to overlay a grid onto any of the views in Raven by
grid lines checking the corresponding box in the layout tab of the side panel. The
spacing of the major and minor grid lines is determined by the spacing of
the major and minor tickmarks on an axis. You can configure which grids
appear in which views by right-clicking on a sound (or Control+clicking
on a Mac) and choosing Configure Grids... from the contextual menu, or
using View> Configure Grids....
Position markers Each view that Raven displays has a horizontal and a vertical position
associated with it, shown by a magenta line, known as a position marker.
You have already seen how the time position marker in a waveform view
indicates the current time during scrolling play (“Filtered play” on
page 9).
When we speak of the “horizontal position marker” we mean
the line that marks the horizontal position, which is a vertical
line.
Notice that when you move the time position of either the waveform or
spectrogram, the time position marker in the other view moves with it.
This is because views that share a dimension (e.g., the time dimension for
waveform and spectrogram views) are by default linked by their position
in that dimension. More detailed information regarding linkage of views
is discussed in “Linking and unlinking views” on page 62.
Centering a position You can move the horizontal or vertical position marker of a view relative
to the window by grabbing it with the mouse and dragging it. To move a
particular point in the data shown in a view to the horizontal or vertical
center of the view pane, place the position marker on the point of interest,
then click the corresponding Center Position button (Figure 3.5). The
position marker and the underlying data will jump to the center of the
50
view pane. Blue labels along the axes show the exact numeric values of the
current horizontal and vertical positions.
WINDOWS
On computers running the Windows operating system,
you can make a view’s horizontal position marker jump
directly to a point by control-clicking at that point.
Figure 3.5. Center position buttons
Figure 3.5. The horizontal and vertical Center Position buttons.
Positions control The horizontal and vertical position of a view control aspects of a view’s
view appearance appearance and behavior in four ways:
and behavior
1. First, a view’s horizontal and vertical position, in combination with the
view’s horizontal and vertical scales, determine what portion of the
view’s data is visible in the window.
2. Second, a view’s position is the point around which zoom operations are
centered.
3. Third, the position of a view in a particular dimension (e.g., time or
frequency) may be used to link that view to any other view that shares
that dimension (“Linking and unlinking views” on page 62).
4. And fourth, during scrolling playback the time position moves through
the data (though the position location stays fixed— see below) and can
be used to control where playback starts (see “Filtered play” on
page 9).
Position vs. position The position of a signal identifies a point in the data along a particular axis,
location such as time or frequency. In contrast, the position location is the point in a
particular view’s panel where the position marker is displayed, i.e., from left
to right or top to bottom. For example, when you drag a time position
marker, you change both the position location (relative to the window)
and the position (relative to the data), since the data do not move with the
marker.2 When you click the Center Position button for an axis, the
51
position marker jumps to the corresponding (horizontal or vertical) center
of the view panel, and the data move with it— i.e., the position location
changes, but the position (relative to the data) does not.
Figure 3.6. Centered pos markers w/gray bkgrd
Figure 3.6. Waveform and spectrogram views with centered position
markers, positioned at the start of the signal (left edge of horizontal
scrollbar) and the lowest frequency (bottom end of vertical scrollbar) of
the spectrogram. The vertical scrollbar refers to the spectrogram view,
because the spectrogram is the active view.
Scale of a view Each view that Raven displays has a horizontal and vertical scale
associated with it. The scale determines the relationship between the
dimensional units shown along that axis (e.g., seconds or kilohertz) of the
view and display units (e.g., pixels, centimeters, or inches) on your
computer screen. The scale at which the entire extent of an axis just fits in
the view pane is called the default scale for that axis. When you first open a
sound file, the time scale of the waveform view is set to the default. When
you first create a spectrogram (information on creating views will be
discussed later in the chapter) the frequency scale is set by default so that
the entire frequency range of the signal fits vertically in the spectrogram
pane.
Setting the scale of The scale and position of the horizontal and vertical axes of any view can
view axes be changed using the zoom controls and scrollbars, as described in the
next section. However, more precise control of scale and position is
available in the Configure View Axes dialog box (Figure 3.7). To display
the Configure View Axes dialog box, choose Configure View Axes... from
the contextual menu for any view or from the View menu for the active
2. The one exception is when you try to drag the position marker beyond the limits of the signal. In that case, the end point of the signal will move with the
marker, and you will be changing the position location but not the position
(which is set to one of its limits already).
52
view. You can enter precise values for the position and scale of the view’s
horizontal and vertical axes. Scale is specified in units per line of the view
(see “Multiple-line views within sound windows” on page 74).
Figure 3.7 Changing time scale
Figure 3.7. The original file (top left) is shown with the default time
scale. Below that (bottom left) is the sound with its time scale changed
to 10 Seconds/Line. The Configure View Axes box is shown (right) with
the edited time scale displayed. You can see the original view shows
more than 20 seconds of the signal, while the bottom view (with the
altered time scale) shows only 10 seconds of the signal across the line.
Changing view At the scale of magnification shown in Figure 3.6, you can’t see individual
scales by zooming cycles of oscillation of the waveform view (top); what you see is the
envelope of the entire signal. In order to see more detail of a signal, you
must adjust the zoom level of the view. In the lower right-hand corner of a
Raven sound window are the zoom controls for the active view (Figure
3.8). Buttons marked with ‘+’ and ‘-’ at the right and bottom ends of the
horizontal and vertical scrollbars respectively increase and decrease
magnification (zoom in and out) around the current position along that
axis.
53
Figure 3.8. Zoom buttons
Figure 3.8. The zoom controls, which apply to the active view. The
Zoom to Selection button is gray if no selection exists in the signal.
Zoom details Each time you click a Zoom In or Zoom Out button, the corresponding axis
of the active view is re-scaled by a factor of 2 (= 1.41). Thus, clicking the
Zoom In or Zoom Out button twice in succession changes the scale by a
factor of 2. To zoom in horizontally on a view, first make sure that view is
active, then move the horizontal position marker to the point where you
want to center the zoom. Click the ‘+’ button at the end of the horizontal
scrollbar, and observe how the display changes. Clicking the ‘-’ button
reverses the change.
Each time that you zoom in or out horizontally or vertically, the
corresponding scroll thumb shortens or lengthens to indicate what
proportion of the signal is visible. (Remember that the scroll thumb is half
the length of the scrollbar when that axis’s scale is set to its default value.)
Zooming in If you click the horizontal ‘+’ (zoom in) button repeatedly the scale will
horizontally eventually be stretched to the point where Raven displays the individual
samples in the waveform as dots (Figure 3.9).
54
Figure 3.9 Zoom to Sample Points
Figure 3.9. A sound window with its waveform view (top) horizontally
zoomed in enough to display individual sample points.
As mentioned before, the spectrogram and waveform views are linked by
their time scale (and time positions.) So, you’ll notice that both the
waveform and spectrogram views zoomed in, even though only one of
them is the active view.
Zooming vertically To zoom vertically in the spectrogram, activate that view, and drag the
vertical position marker to the point where you want to center the zoom.
Click the ‘+’ or ‘-’ button at the bottom of the vertical axis to increase or
decrease magnification around the vertical position marker.
Notice that changing the vertical position or scale of the spectrogram has
no effect on the waveform. The vertical dimensions of the views are
different (amplitude in the waveform, frequency in the spectrogram), so
the vertical scales of the views cannot be linked.
Zoom to all To reset the horizontal or vertical scale of a view to display the entire axis,
click on the corresponding Zoom To All button, marked with a horizontal
or vertical I-beam icon (Figure 3.8, page 54).
Zoom to selection The Zoom To Selection button rescales the active view so that the active
selection fills the view frame. If there is no active selection, the Zoom To
Selection button is not available. In the spectrogram view, select a
rectangular area to enlarge by clicking and dragging the mouse pointer.
Then click on the Zoom To Selection button in the lower right corner of the
sound window (Figure 3.8, page 54).
55
View Types
The six main view Waveform view
types When you open a sound in Raven, the top view in a sound window is a
waveform view by default. This view displays an oscillogram which is a
graph of the sound showing amplitude versus time.
Spectrogram view
The second view shown by default in a sound window, below the
waveform view, is a spectrogram view. It shows time on the horizontal
axis, frequency on the vertical axis, and relative power at each point as a
color (grayscale by default.)
Spectrogram Slice view
A spectrogram slice view is a plot of relative intensity versus frequency at
a particular point in time within a signal. A spectrogram slice represents a
vertical cross section through a spectrogram at a single time, but is rotated
90° so that the frequency axis is horizontal. In fact, a spectrogram is built
of a series of spectrogram slices stacked side by side (with their frequency
axes running vertically).
Where a spectrogram view shows a series of slices at successive points in
time, and represents power at each frequency by a color value, a
spectrogram slice view shows only one slice and represents power at each
frequency on a line graph. The relationship between spectrogram and
spectrogram slice views is explained in more depth in Chapter 5,
“Spectrographic Analysis”.
Selection Spectrum view
These show the average spectrum of a sound over the time interval
defined by a user-made active selection. More information about
selections and how they affect this view type can be found in “Selection
spectrum views” in Chapter 5 (page 139).
Beamogram View
Beamforming provides a means for determining a sound source's likely
bearing relative to a multi-channel array of sensors. The beamogram view
creates a plot of potential bearing angles versus time. Areas of higher
energy in the beamogram view correspond to greater likelihood that
specific angles represent actual sound source bearings. More information
about determining source directions using the beamogram view can be
found in “Bearing analysis: Beamforming, Beamogram View, and Max
Bearing Measurement” in Chapter 8 (page 213).
56
Beamogram Slice view
The beamogram slice view represents a vertical cross section of the
beamogram view, showing the potential bearing angles at a specific point
in time. Potential bearing angles are plotted on the horizontal axis, with
vertical values corresponding to the likelihood that a sound is coming
from the corresponding direction. More information regarding the
beamogram slice view and its relation to the beamogram view can be
found in “Bearing analysis: Beamforming, Beamogram View, and Max
Bearing Measurement” in Chapter 8 (page 213).
Color Bar View
A color bar view serves as a key to the current spectrogram color scheme.
It displays the spectrum of colors in the color scheme along with their
associated intensity values. While this view contains no new information
about the sound, it can be helpful for visually interpreting spectrograms.
Making a new view New views are created by using the appropriate buttons located on the
view toolbar (Figure 3.10).
Figure 3.10 New view buttons on the control toolbar
Figure 3.10. The new view buttons on the view toolbar.
Open a sound file like ChestnutSidedWarbler.aif. The sound window
shows a spectrogram and waveform view by default. To create a new
view, like a spectrogram slice view, simply click on the New Spectrogram
Slice button in the view toolbar. Alternately, you can choose View > New >
Spectrogram Slice View..., or type <Ctrl-L> (Windows, Linux) or
<Command-L> (Mac OS). The Configure New Spectrogram Slice View
dialog box should appear (Figure 3.11.).
57
am Slice View dialog box
Figure 3.11. The Configure New Spectrogram Slice View dialog box.
This dialog box lets you specify various parameters that affect the
appearance of the spectrogram. For now, don’t worry about what these
parameters mean; they are explained in detail in “Configuring
spectrographic views” in Chapter 5 (page 112). Click OK to accept the
default settings and compute the spectrogram slice view.
The sound window is re-drawn with an additional view pane, containing
the new spectrogram slice view. If the time position marker of the
waveform or spectrogram view is at the very beginning of the data (where
Raven puts it by default when opening a new signal) the spectrogram slice
view does not display any data (Figure 3.12).
58
Figure 3.12 New Spectrogram Slice View
Figure 3.12. This sound window contains the usual waveform and
spectrogram views, but it also contains a blank spectrogram slice view
as well. The spectrogram slice view displays no data when the time
position marker of the spectrogram and waveform view is located at the
start of the data (all the way to the left).
Showing data in a To see valid data, drag the time position marker to the right in either the
new view waveform or spectrogram view (Figure 3.13). (For a complete explanation
of why no data are visible in the spectrogram slice view when the position
marker is at the start of the signal, see Chapter 5, “Spectrographic
Analysis”.)
59
Figure 3.13 Spectrogram Slice view with data
Figure 3.13. The same sound window as shown in Figure 3.12 but with
the time position marker (the vertical magenta line) moved to a position
in the waveform and spectrogram data. The spectrogram slice view
now displays the data from the time designated by the time position
marker.
Selected view Remember that the active view in a sound window is marked by the
dictates toolbar highlighted view selection button to the left of the view (for example, the
content spectrogram view is active in Figure 3.13). The activated contents of the
toolbars are dictated by which view is active in the current sound window.
For example, if a spectrogram view is active, the brightness and contrast
sliders in the spectrogram toolbar become active (Figure 2.12 on page 37).
If any of the other views are active, the brightness and contrast sliders are
grayed out because they have no relevance in those cases.
The zoom window The zoom window is a separate window on the Raven desktop that can be
used to view selections from any open sound. Once a zoom window is
opened, it displays a magnified view of the active selection of the active
sound. If the zoom window is the active window, it will display the active
selection from the previously active sound. The zoom window
automatically updates as you move through selections in a sound, making
it particularly useful for browsing details of selections without having to
manually zoom in on each one before advancing.
To open a zoom window, choose Zoom Window from the Window menu.
You can configure the views or change display settings in the zoom
window through the context menu or side panel in the same way as other
sound windows. These settings are applied specifically to the zoom
window, and override the display settings of the original sound when
displaying selections. For multi-channel files the zoom window only
60
displays the channels containing the active selection; however, you can
choose to add other channels from the side panel. Figure 3.14 shows the
default zoom window with the parent sound window for a selection.
Figure 3.14Zoom Window
Figure 3.14. The zoom window displays a magnified view of the active
selection, showing only those channels that contain the selection.
When the zoom window is active the cursor will change from its current
setting to a play cursor, allowing you to play the selection simply by
clicking in the zoom window. This is the case for both the create selection
and activate selection modes. However, the Grab and Scroll mode retains
its normal function in the zoom window.
When a workspace is saved, the zoom window’s visibility and settings will
be saved as well. However, if the zoom window is active when the
workspace is saved, its contents may not be fully restored. If another
sound window is active, however, the zoom window will restore
accordingly.
61
Linking and unlinking views
Two views that share a dimension— such as time or frequency— can have
their positions and scales either linked or unlinked by that dimension. When
the positions of two views are linked, moving the position of either one
(by moving the corresponding scroll thumb, or by moving the position
marker within the window) causes the positions of both views to move
together. When the scales of two views are linked, changing the scale of
either one (using the zoom buttons) causes both views to be re-scaled. In
the examples you’ve seen so far, waveform and spectrogram views have
been linked by time position and in time scale.
When you open or record a signal and create multiple views, each view is
initially linked in position and scale to every other view that shares its
dimensions. For example, waveforms and spectrograms are initially all
linked to each other in time position and time scale, and spectrogram and
spectrogram slice views are linked in frequency position and frequency
scale. Spectrograms and waveforms cannot be linked in frequency because
waveforms have no frequency dimension.
Unlinked views In some situations, it can be useful to work with views that are unlinked
from each other in particular ways. For example, Figure 3.15 shows two
spectrogram views of a signal that is 53 seconds long. The views are linked
in time position but unlinked in time scale. Unlinking the time scales
makes it possible to view simultaneously an eight-second portion segment
of the signal (upper view), and a close-up of a two-second segment
contained within it (lower view). Since the two views remain linked in
time position, the position markers provide a reference point to show
where the close-up view is located within the entire signal. As either view
is scrolled horizontally through the signal, the linked position markers
identify the same point in time in both views.
62
Figure 3.15. Two linked spgs, diff time scales
Figure 3.15. Two spectrogram views of the file “BlackCappedVireo.aif”
that are linked in time position, but unlinked in time scale. The upper
spectrogram shows 7.8 seconds of the signal; the lower spectrogram is
a close-up of 2.0 seconds. The highlighted rectangle identifies the
same selection in both views. The time position markers in both
spectrograms identify the same time, because the time positions are
linked.
Unlinking views To create two spectrograms that are unlinked in time scale as discussed in
through the the above example (Figure 3.15):
contextual menu 1. Open a signal.
2. Make a second spectrogram by clicking the New Spectrogram button
(Figure 3.10, page 57), or typing <Ctrl-G> (Windows, Linux) or
<Command-G> (Mac OS).
Place the mouse pointer on the second spectrogram, activate the
contextual menu, and choose Unlink View > Time Scale.
Now try using the zoom buttons to change the time scale of the first
spectrogram. Notice that the time scales of the waveform and the first
spectrogram change together, but the time scale of the second spectrogram
does not change. If you activate the second spectrogram, and zoom in
time, its time scale changes, while the time scales of the waveform and the
first spectrogram remain unchanged. If you scroll any of the three views,
their position markers will move in synchrony, because all three are still
linked by time position.
63
Restoring or
creating new
linkages through
the linkage tab
To restore linkages between views that have been unlinked, or to create
linkages between views that have never been linked, activate the Linkage
tab in the Raven side panel. (If the side panel is not visible, click on the
right-pointing triangle at the top (Windows, Linux) or bottom (Mac OS) of
the separator bar at the left edge of the Raven window.)
At the top of the Linkage tab, there is a drop-down menu showing all of
the linkable properties of the current views. Click on this menu, and
choose Time Scale. If you created two spectrograms and unlinked them as
described above, the tab will display three folder icons representing
linkage groups, with names like “Group 1”, “Signal 1 Default”, and “New”
(Figure 3.16). (The second group may be called “Signal 2 default”, “Signal
3 default”, etc., depending on how many signals you’ve opened in the
current Raven session. If more than one signal is open, you will see a
default linkage group for each signal.)
Figure 3.16 . Linkage tab.
Figure 3.16. The Linkage tab, showing the linkage groups for the Time
Scale property corresponding to Figure 3.15
A linkage group is a set of views that are linked to each other by a
particular property (in this case, Time Scale). Views that are in different
linkage groups are unlinked from each other in that property.
Open the first two folders by double-clicking on them (or single-clicking
the icon to the left of each folder icon). In the present case, the linkage
group called “Sound 1 Default” contains the views called “Sound 1
Waveform 1” and “Sound 1 Spectrogram 1”; Group 1 contains “Sound 1
Spectrogram 2”. The active view is highlighted, and you can activate a
different view by clicking on its name.
To link two views that are presently unlinked, drag the icon for one of the
views to the linkage group that contains the other. In the case shown in
Figure 3.16, you can unlink the waveform from Spectrogram 1 and link it
64
to Spectrogram 2 by dragging its icon from the Default folder to the Group
1 folder.
When you move a view from one linkage group to another, the view is
immediately redrawn if necessary so that its image has the correct scale,
position, or other linkable property. If a linkage group contains only a
single view, and you drag that view to a different linkage group, the
original (now empty) group disappears.
Creating a new To create a new linkage group select one of the views listed by clicking on
linkage group it, and drag it onto the folder labeled “New”. A new linkage group will
appear in the list, containing the view that you moved.
Creating new linked To create a new spectrogram, spectrogram slice, or selection spectrum
views view that is linked by spectrogram parameters to an existing
spectrographic view, choose New > Similar Spectrogram View, New >
Similar Spectrogram Slice View, or New > Similar Selection Spectrum View
from the contextual menu for any spectrographic view, or from the View
menu for the active spectrographic view.
Other linkable In addition to its position and scale for each of its dimensions, every view
properties has four other properties by which it can be linked to other views:
•Position Location: Moving the location of the time or frequency position
bar in a view will also move the corresponding position bar in other
views that are linked by Position Location.
•Color Scheme: Color schemes include color specifications for most of
the graphical elements shown in a view (e.g., axes, backgrounds,
selection boundaries), as discussed later in this chapter. Adjusting
brightness and contrast settings for a spectrogram adjusts the settings for
all spectrograms linked to it by Color Scheme. By default all views of a
signal are linked to each other by Color Scheme.
•Channel Visibilities: In multi-channel signals (see Chapter 8, “Multichannel Sounds”), hiding or showing particular channels of a view will
also hide or show those channels in other views that are linked by
Channel Visibility.
•Number of Lines: Changing the number of lines displayed for a view
(see “Multiple-line views within sound windows” on page 74)
automatically changes the number of lines displayed for any other views
that are linked by Number of Lines. By default, all views in a signal that
display a time axis are linked to each other by Number of Lines.
Additionally, spectrogram, spectrogram slice, and selection spectrum
views can be linked by spectrogram parameters. If two views are linked
65
by spectrogram parameters, then changing the parameters for either view
automatically changes both views.
Views that share a linkable property can be linked even if they
are in different signal widows. This can be useful for example
in setting views of different signals to the same scale.
Linkage between
spectrogram and
spectrogram slice
views
When you request a spectrogram slice view, Raven actually calculates an
entire spectrogram of the signal, but only displays one slice at a time. A
spectrogram slice view thus has a “hidden” time axis. Which slice of the
underlying spectrogram is displayed is determined by the view’s time
position. By default, the time position of a spectrogram slice view is linked
to the time positions of waveform and spectrogram views. As you move
the time position marker of a spectrogram view, any spectrogram slice
view that is linked to it is continually updated to show the cross-section of
the spectrogram at the time position marker.
Even though a spectrogram slice view has a time position, it
does not have a time scale, since the time axis of the underlying spectrogram is not displayed.
By default, new spectrogram and spectrogram slice views are also linked
to each other by frequency position and frequency scale. When you move
the frequency position of a spectrogram, the frequency position of any
linked spectrogram slice views move as well, and vice versa. Similarly,
when you zoom the frequency scale of any spectrogram slice view, the
frequency scale of any linked spectrogram zooms as well. It is important
to remember that the frequency scales of spectrogram and spectrogram
slice views are rotated 90° from each other.
Controlling how views are displayed
Using tools and commands in the side panel’s Layout tab and in the View
menu, you can rearrange the order of views and hide or show the nondata components of views such as the axes and position markers.
Another way to hide a view is to activate that view and choose Hide View
from the View menu, or from that view’s contextual menu.
Deleting a view To delete a view, activate the view and choose Delete View from its
contextual menu or from the View menu. When you delete a view (as
opposed to hiding it), the memory occupied by that view is released for
reuse. Re-displaying a hidden view occurs instantly, whereas re-
66
displaying a view that was deleted requires that it be recalculated, which
may take a significant amount of time.
When you activate a contextual menu in a view pane, viewspecific commands in the menu apply to whichever view the
mouse pointer is on when the menu is activated, irrespective
of whether or not that view is active.
Changing the order To rearrange the order in which views are displayed in the sound window,
in which views are you can drag and drop view names in the Layout tab.
displayed
You can also move a view up or down within the sound window by
choosing Move View Up or Move View Down from the View menu or the
view’s contextual menu.
Hiding and showing The bottom pane of the side panel’s Layout tab contains a list of window
window components that can either be displayed or hidden. You can display or hide
components each component of the active window by checking or unchecking its
checkbox. Figure 3.17. shows a sound window containing two views, with
all components hidden.
Figure 3.17.. All components hidden.
Figure 3.17. A sound window with all components hidden.
By default, all window components are displayed except for line titles,
which identify the view type and number of each view (Figure 3.18.). If the
sound has multiple channels (see Chapter 8, “Multi-channel Sounds”) or a
view has multiple lines, channel and line numbers are displayed in line
titles as well.
67
Figure 3.18. . View titles.
Figure 3.18. A sound window showing a waveform and a spectrogram
view, with line titles displayed.
Changing the Appearance of a Sound Window
Spectrogram When a spectrogram view is active, the brightness and contrast controls
brightness and on Raven’s spectrogram toolbar become active too (Figure 2.12 on page
contrast 37).
If your spectrogram looks too dark or light, or if it’s hard to pick the signal
out of the background, move the brightness and contrast sliders to achieve
the desired appearance of the spectrogram. The brightness control adjusts
the overall darkness of the spectrogram: for a grayscale spectrogram (the
default), sliding the control to the right lightens the display.
The contrast control adjusts the number of different color (by default,
grayscale) values that are shown in the spectrogram. In a grayscale
spectrogram, moving the contrast slider all the way to the right makes the
display black and white: all values below some threshold are assigned to
white and the rest become black. In this case, the threshold between black
and white is determined by the brightness control. With the contrast
control all the way to the left Raven displays up to 200 shades of gray.
The box next to each control tells you what percent contrast or brightness
you’ve set. If you prefer, you can type a percentage number into a box
instead of moving the slider.
In addition to the toolbar controls, you can also configure spectrogram
brightness and contrast by choosing Configure Brightness and Contrast...
from a view’s contextual menu. In addition to a set of brightness and
68
contrast sliders, the resulting dialog allows you to alter these parameters
by specifying floor and ceiling values for the spectrogram and displays a
plot illustrating how colors are assigned to each value in a spectrogram.
The floor and ceiling threshold parameters determine the power levels
beyond which all values are assigned the same color. In the case of a
grayscale spectrogram, values below the floor threshold are all displayed
as white, and values above the ceiling threshold are all displayed as black.
For more information on the relationship between these threshold
parameters and corresponding brightness and contrast values, see
“Brightness and contrast” in Chapter 5 (page 130).
Figure 3.19Brightness and Contrast Dialog
Figure 3.19. The configure Brightness and Contrast dialog
Color schemes of Each view in a sound window has a particular color scheme. A color
sound windows scheme consists of a spectrogram colormap and particular colors for items
such as plottable measurements, active and inactive selection borders and
fills, and the waveform line.
69
A spectrogram colormap defines the relationship between power values in
the spectrogram and colors used to display them. Raven provides six
predefined colormaps: Grayscale, Hot, Cool, Standard Gamma II3, Bone,
and Copper (Figure 3.20.). You can also create customized colormaps as
text files as described in “Creating customized colormaps” in Chapter 11
(page 296). To change the colormap of a spectrogram, choose Color
Scheme > Colormap-name from the spectrogram’s context menu, or from
the View menu when the spectrogram is active. You can reverse the
spectrogram colormap by choosing Color Scheme > Reverse Color Map
from the spectrogram’s context menu, or from the View menu when the
spectrogram is active. To display the spectrum of colors in the color map
along with their associated intensity levels, you can add a color bar view
to the sound window by clicking on the icon in the view toolbar or
through the View > New menu.
Views in a sound window are by default linked to each other
by color scheme. If you change the colormap or any other
item in a view’s color scheme, the affected colors will also
change in any other views that are linked to it by color
scheme.
3. The Standard Gamma II color map is included in Raven courtesy of Research
Systems, Inc., a Kodak Company.
70
Figure 3.20.. Colormap examples
Figure 3.20. Four spectrogram views of the signal
“CassinsKingbird.wav”, using four different, predefined colormaps. (a)
Hot. (b) Standard Gamma II. (c) Bone. (d) Copper.
You can reverse the spectrogram colormap by choosing Color Scheme >
Reverse Color Map from the spectrogram’s context menu, or from the
View menu when the spectrogram is active.
In general, if the background of a spectrogram is white, then
light colors represent low power. If the background is black,
the dark colors of a colormap represent the low power.
Editing color In addition to selecting spectrogram colormaps, you can also edit a view’s
schemes color scheme by choosing Color Scheme > Edit... from the view’s context
menu, or from the View menu when the view is active. More information
on customizing color schemes can be found in “Editing color schemes” in
Sound Window You can save information about the layout of a sound window in a window
Presets preset, which you can later apply to another sound window (or to the same
71
window). A window preset includes information about the size of a
window, what views are displayed, whether or not the selection table is
showing, and other properties (Table 3.1).
72
Table 3.1. Window properties saved in window presets.
window size
view types, parameters, color schemes,
positions, scales, sizes
spectrogram parameters
spectrogram brightness and contrast
selection table visibility and size
measurement list and annotation columns
selection label configuration
selection label visibility
axis visibility
major and minor grid visibility
line title visibility
position marker visibility
view selection button visibility
scrollbar visibility
To save a window preset, first configure the properties listed in Table 3.1
the way you want to save them, then choose View > Window Preset >
Save As... . When the Save Sound Window Preset dialog box appears,
enter a name for the preset, and click OK.
Window presets must be saved in the folder Presets/
Sound Window/ within the Raven program folder. You can
also create additional folders within the Sound Window
folder by clicking on the New Folder icon within the Save
Sound Window dialog. These folders will appear as submenus in the Window Preset menu, with each submenu listing the presets in the corresponding folder.
To apply a window preset that’s already been saved to the active sound
window, choose View > Window Preset > Preset-name. When you apply a
saved preset, the active window is redrawn with the properties specified
in the preset, and any additional views are calculated as needed. If any of
the open selection tables contain annotation columns, Raven will display a
dialog where you can choose to keep, replace, or merge these columns
with those in the preset.
73
You can specify a window preset to apply when you first open a sound file
in the Configure New Sound Window dialog box (Figure 1.3, page 4). You
can change the name of the default window preset by editing the Raven
preferences file, as described in “About Raven preferences” in Chapter 11
(page 299).
If you make changes to the properties of a window to which you’ve
applied a preset, you can save the changes to the current preset by
choosing View > Window Preset > Save “Preset-name”.
Window presets take precedence over other presets that
affect the appearance of the contents of a window. Thus, if
you apply a window preset, spectrogram parameters, the
measurement list, and the color scheme are all set as defined
in the specified window preset, even if they were previously
set to different values by choosing spectrogram, measurement, or color scheme presets.
Multiple-line views Raven can display views that have a time axis (i.e., waveforms and
within sound spectrograms) across multiple lines (Figure 3.21). The values shown on the
windows time axis of a multiline view start at the left end of the top line, go to the
right end of that line and continue on from left to right on subsequent
lines, like lines of text on a page.
74
Figure 3.21. Multi-line spg.
Figure 3.21. A spectrogram view with three lines. Note the continuity
of the time axis from one line to the next.
On the side panel, the Layout tab’s Lines box shows you how many lines
there are for the selected view. To change the number of lines, select a
waveform or spectrogram view, change the number in the box, and press
<Enter>. By default, all waveform and spectrogram views in a sound
window are linked to each other in the number of lines displayed. You can
unlink the number of lines property as described in “Linking and
unlinking views” on page 62.
The scale of the time axis doesn’t change when you change
the number of lines displayed. To rescale the axis so that the
view fills the available lines, click on the horizontal Zoom-toAll button at the right end of the horizontal scrollbar.
Grouping views In sound windows with two or more views and two or more lines per
with multiple lines view, you can change the way the lines and views are grouped in the
window. By default Raven displays all the lines for one view, followed by
the lines for the next; this is called grouping by view (Figure 3.22). The
alternative is to see all the first lines for all views, followed by all the
75
second lines for all views, and so on; this is called grouping by time (Figure
3.23).
Figure 3.22. VL grouping.
Figure 3.22. A sound window containing two views, each with two
lines, grouped by view. All lines of the waveform view are in the top
group, all lines of the spectrogram view are in the next group.
76
Figure 3.23. LV grouping.
Figure 3.23. The same two views shown in Figure 3.22, but grouped
by time. The first lines of both views are in the top group; the second
lines of both views are in the next group.
To change the grouping, choose View or Time from the pull-down menu
labeled “Group By”, in the side panel’s Layout tab.
The horizontal separator bar that Raven displays between
top-level groupings (i.e., between views in View grouping or
between lines in Time grouping) is wider than the bar used
between second-level groups.
Working with more Raven lets you work with more than one sound at a time. When more than
than one sound one sound window is open, the tools in Raven’s toolbar (e.g., New View
buttons, brightness, contrast, and playback controls) and in the side
panel’s Layout and Playback tabs always refer to the active window (the one
window in which the title bar is colored). The side panel’s Linkage tab
shows views for all sounds that are open, since Raven lets you link views
of different sound windows to each other. You can activate a sound
window by clicking anywhere in it with the mouse. You can also activate a
77
window by choosing its name from the Window menu, or by clicking on
the name of one of its views in the side panel’s Linkage tab.
Tile/Cascade When you have multiple sound windows open, Raven can arrange them
windows neatly for you on the desktop. If you choose Window > Tile Windows,
Raven resizes and repositions all of the windows so that they are all the
same size and fill the desktop (as nearly as possible) without overlapping
each other. Choosing Window > Tile Windows Horizontally arranges the
windows in a row such that they are all the same width and fill the
desktop. Similarly, choosing Window > Tile Windows Vertically arranges
the windows in a column such that they are all the same height and fill the
width of the desktop. See Figure 3.24 for examples of these tiling options.
If you choose Window > Cascade Windows, Raven resizes and repositions
all of the windows so that they are all the same size and as large as
possible, overlapping so that their title bars are all visible (Figure 3.25).
78
Figure 3.24 Tile Windows
Figure 3.24. Examples of tiled sound windows (a) 2 to 3 sounds will be
tiled in one column (b) 4 to 8 sounds will be tiled in two columns and (c)
9 to 12 sounds will be tiled in three columns. (d) 4 windows tiled
horizontally. (e) 4 windows tiled vertically
79
Figure 3.25 Cascade Windows
Figure 3.25. An example of cascaded sound windows. Here, the
windows are stacked on each other with all title bars visible.
Apply current To apply all of the active window’s layout to all currently open windows,
window layout to all choose View > Apply To All. Window layout includes all of the information
windows that would be stored in a sound window preset, including window size,
view visibility and scale, spectrogram parameters, and spectrogram
brightness and contrast. A complete list of window properties affected by
the Apply To All command is given in Table 3.1.
80
Chapter 4
Signal Acquisition (Recording)
About this chapter
This chapter describes how to acquire (digitize) signals with Raven, and
assumes that you are familiar with basic Raven concepts and operations as
described in earlier chapters.
In order to make good recordings with Raven, you should be acquainted
with the basic concepts of digital sound representation discussed in
Appendix A, “Digital Representation of Sound”. Failure to understand
concepts such as sample rate and sample size can lead to digitized signals
that are poor in quality or distorted in various ways.
In this chapter you’ll learn how to:
•choose whether to record to memory, to a file, or to a file sequence
•create and configure a recorder window
•create and work with real-time signal views in a recorder window
Overview of signal acquisition
Raven acquires (records) audio signals via recorders. A recorder is
represented on the Raven desktop by a recorder window that can display
one or more views of a signal in real time as the signal is being acquired. A
recorder window is like any other Raven sound window, except that it has
additional controls for starting and stopping recording and cannot be
edited.
Recording modes Raven provides three different recording modes, which differ in where the
signal is stored as it’s recorded.
•Record to memory: When recording to memory, audio data are not
automatically saved to disk files, but reside only in a portion of Raven’s
temporary memory allocated as a recording buffer. You can specify the
size of the buffer (in seconds) when you configure a recorder. During
recording, audio data are stored in the buffer until the buffer is full, and
then the oldest data are continually discarded to make room for new
data being acquired. For example, if the buffer size is 30 seconds (the
default), the buffer will always contain the most recent 30 seconds of
data during recording; data recorded more than 30 seconds ago are lost.
While a signal is in the recording buffer, you can manually save it to a
disk file, or save selections from it. Recording to memory is useful for
exploratory browsing through recordings when you don’t need to save
81
Chapter 4: Signal Acquisition (Recording)
the data to files, or when the signals you’re interested in saving are
sparsely distributed among sections that you don’t want to save.
•Record to file: When recording to a single file, Raven saves audio data
directly to a disk file for a specified amount of time, then stops. You can
also manually stop recording to disk before the specified time has
finished.
•Record to file sequence: When recording to a file sequence, Raven saves
audio data continuously to a series of files, each of which is a specified
size. Successive files in a sequence are named according to a template
that you specify, which can incorporate a sequence number or a
timestamp. Raven also saves a list file, which can be used to open the file
sequence as a single continuous signal.
Creating a recorder You create a recorder by specifying a set of configuration parameters in
the Configure New Recorder dialog box (Figure 4.1), which is invoked by
clicking on the Record button on the Raven toolbar (identified by the
microphone icon, left) or by choosing File > New Recorder....
Figure 4.1. New Recorder dialog.
Figure 4.1. The Input tab of the Configure New Recorder dialog box.
You use the Record To drop-down list in this dialog to choose whether to
record to Memory, a File, or a File Sequence. Depending on which
recording destination you choose, different tabs appear in the dialog box.
The following sections discuss the configuration options on each tab.
Recording Input
The Input tab (Figure 4.1) allows you to choose and configure an audio
input device. The Input tab is displayed for all three recording modes.
82
Device Click on the Device drop-down list to see a list of all available audio input
devices. By default, the operating system’s default audio input device is
selected. The name that is shown for this device will be the name of the
default audio input device (typically a built-in sound card) for your
computer. The audio input device obtains its input from a recording source
that you choose using controls supplied by your computer’s operating
system. Appendix C, “Configuring Audio Input” explains how to
configure your operating system’s controls to choose a recording source
(e.g., the computer’s built-in CD player, a microphone, or line input jack)
for audio input.
Each audio input device is characterized by a set of available sample rates
at which signals can be recorded, the number of channels that can be
recorded, and a set of sample formats. Each of these parameters is
discussed below. Some devices may also have additional parameters that
can be configured.
Raven now includes support for multi-channel recording
using both NI-DAQ and ASIO audio input devices. For more
information about using this feature, please see Appendix C,
“Configuring Audio Input”.
Sample Rate The Sample Rate drop-down list displays all of the sample rates available
for the selected input device. Typically, the sample rates available are
determined by the underlying sound input hardware. The Sample Rate
parameter specifies the rate at which an analog electrical signal
representing a sound is sampled to create the digital signal. Be sure to
choose a sample rate that is at least twice as high as the highest frequency
in the recording you want to acquire, in order to avoid signal distortion
due to aliasing. See Appendix A, “Digital Representation of Sound” for a
discussion of aliasing and the trade-offs involved in choice of sample rate.
Channels The default audio input device on most computers provides for one(mono) or two-channel (stereo) acquisition. For signals that are acquired in
stereo, the Left and Right channels are identified within Raven as Channel
1 and 2, respectively. To acquire only one channel, uncheck either the Left
or Right checkbox. The remaining channel will be acquired as Channel 1.
If you selected a device that supports input of more than two channels,
each channel is represented by a numbered checkbox. If you disable input
of some channels, data from the channels that are recorded will be
represented in the resulting acquired signal in channel-number order, and
renumbered sequentially beginning at Channel 1. For example, if you are
recording from a 16-channel input device, and you select only input
channels 1, 5, 7, and 12, then data from these channels will be represented
in channels 1, 2, 3, and 4 respectively of the acquired signal.
83
Sample Format The Sample Format drop-down list shows all of the sample formats
available for the selected virtual recording device. The default input
device on most computers supports two formats, 8-bit and 16-bit signed
PCM (Pulse Code Modulation). Some sound cards will support 24-bit
signed PCM. A larger sample size (more bits) results in a better quality
digitized signal (higher signal-to-noise ratio), but may also increase the
storage requirements for signals saved to disk files. See Appendix A,
“Digital Representation of Sound” for a discussion of the trade-offs
involved in choice of sample size.
Raven always represents signals internally with 32-bit floating point
samples. As a result, Raven accurately stores signals recorded with signal
sizes up to 24 bits. If you choose an input sample size greater than 24 bits,
the input values may be truncated, resulting in a distorted signal. (for
more on floating point values and truncation, see the footnote under
“Sample Size” on page 87).
Recording Display
The Display tab (Figure 4.2) allows you to configure the recorder window,
and is displayed for all three recording modes.
Figure 4.2 . Recorder Display tab.
Figure 4.2. The Display tab of the Configure New Recorder dialog box.
Buffer Size The buffer size determines how much of the signal is retained in memory
and can be visible in the recorder window during signal acquisition. You
84
can specify buffer size either in seconds or minutes. You can use the dropdown list to choose time units of seconds or minutes. The default buffer
size is 30 seconds.
During recording, audio data remain in memory for the duration of the
buffer, and are then discarded to make room for new incoming data. If
you’re recording to memory, data that are discarded from the buffer are
lost permanently. When recording to a file or a file sequence, data that are
gone from the buffer can be retrieved by opening the file(s) that were
saved during the acquisition.
Update Rate The update rate determines how many times per second views in the
recorder window are redrawn during signal acquisition. The default rate
is 10 Hz. There is a trade-off between the perceived smoothness of the
scrolling real-time display and Raven’s responsiveness to user interaction.
If the update rate is too low, scrolling views move in visibly discrete steps,
resulting in a jittery display. If the update rate is too high, Raven will be
slow to respond to user actions such as mouse clicks or keyboard
commands. In practice, the best rate will depend on the size of the
recorder window, the number and type of views it contains, and the speed
of the computer. The default update rate of 10 Hz provides a good balance
between display smoothness and performance on most computers. You
may need to experiment to determine what update rate provides the best
performance on your computer. See “Reconfiguring an existing recorder”
on page 98 to learn how to change the update rate once a recorder is
running.
Latency The Latency value on the Display tab specifies the difference, for a running
recorder, between the current time and the active view’s time axis position
(by default, the right-hand edge of the window). Specifying a display
latency greater than zero introduces a delay in the display of data in realtime views.
By specifying a display latency greater than zero, you give Raven some
additional time to compute a view and build the corresponding image
before it needs to be displayed. A small display latency (e.g., 0.1 or 0.2
seconds) often eliminates the flickering grey band that is sometimes
visible at the right-hand edge of some real-time views, while only
resulting in a barely-perceptible delay in the display of real-time data.
Non-zero display latencies also result in smoother updates to real-time
spectrogram slice views.
Display latency vs. processing latency
When a recorder is running, you may observe a flickering gray band at the
right-hand edge of some real-time views. This band is more evident in
recorder windows that are displayed at large sizes or that contain many
views, when the time axis is zoomed in, and on slower computers. This
band represents the view’s processing latency, which is the difference
between the current time (by default the right-hand edge of the window)
and the last time for which Raven has displayed data in the view. The
85
processing latency is always greater than zero because it takes some time
for Raven to build the visual image of the data to be displayed in the view.
Real-time views scroll at a fixed rate, irrespective of the rate at which
Raven builds images for display. When a view scrolls and no new image
data are available for display, Raven draws a gray band to fill in the time
interval for which the image is not yet available. As soon as the image is
computed, it replaces the gray band. Processing latency is typically longer
for spectrogram or spectrogram slice views than for waveforms, because
spectrograms require more calculation than waveforms. The processing
latency (hence the width of the gray band) varies from moment to moment
for each view as Raven rapidly switches among the various tasks involved
in updating the views in the recorder window. When the processing
latency is small relative to the time scale shown in a view, the flickering
gray band disappears. When it is larger, the band becomes more
prominent.
Window preset The Window Preset drop-down list on the Display tab shows the name of
the window preset that will be applied to the recorder window when it is
first created. Window presets define what views are visible in a window,
as well as window size, selection mode, and other properties. You can
choose from all of the window presets that are currently defined by
clicking on the drop-down menu icon to the right of the Window Preset
field. Window presets are discussed further in “Sound Window Presets”
File Format
The File Format tab appears only if you choose Record to File or Record to
File Sequence.
86
Figure 4.3. File Format tab.
Figure 4.3. The File Format tab of the Configure New Recorder dialog
box.
File Format Raven can acquire audio data directly to files saved either in AIFF (Audio
Interchange File Format, filename extension .aif) or in WAVE format
(filename extension .wav). WAVE files can be opened by most other
programs that work with audio data. AIFF files can be opened by most
Macintosh programs that work with audio data, and some programs on
other platforms. Raven can open either type of file.
Sample Size The Sample Size parameter specifies the “bit depth” of the samples saved
in the file(s) created during signal acquisition. Raven always represents
signals internally with 32-bit floating point samples. Therefore, values
recorded using an input sample size greater than 24 bits may be truncated,
resulting in distorted signal.1See Appendix A, “Digital Representation of
Sound” for a discussion of the trade-offs involved in choice of sample size.
1. In general, individual samples in a signal of sample size n can have 2n possible
values, from -2n-1 to 2n-1-1. Thus, in general a 32-bit representation has
4,294,967,296 (=232) possible sample values, from -2,147,483,648 to
+2,147,483,648. However, by using a floating point representation, Raven
stores each value as a combination of a 24 bit integer and an 8 bit exponent.
This allows values across a much greater range to be stored, but limits the precision to which larger values can be represented. As such, signals recorded
using an input sample size greater than 24 bits will be truncated to include
only 24 bits of precision, resulting in increased distortion for large values.
87
File Size The size of the file(s) to be saved can be specified either in time units
(seconds or minutes), storage units (kilobytes or megabytes), or sample
frames. A sample frame consists of one sample for each channel in the
acquired signal. For example, if you specify a file size of 100,000 sample
frames for a signal acquired at a sample rate of 22,050 Hz, each acquired
file would be 4.535 seconds long (4.535 = 100,000 sample frames / 22,050
sample frames per second). If only a single channel is acquired, the file
would contain 100,000 samples; a stereo signal would contain twice that
many. If the files are saved with a sample size of 16 bits (2 bytes), a singlechannel file would contain 200,000 bytes of data; a stereo file would
contain 400,000 bytes.2
File naming
If you choose Record to File or Record to File Sequence, the File Name or
File Names tab appears (note the plural tab label when recording to file
sequences). The fields on these two tabs allow you to specify where to save
acquired files and how to name them. The only difference between the File
Name and File Names tabs is that the latter allows you to specify a name
for a list file, which is required when recording to file sequences. See
“Recording to a file sequence” on page 92 for more about list files. This
section explains how to specify file names for both single-file and file
sequence recording.
2. The actual total size of AIFF or WAVE files is slightly larger than the number of
bytes of audio data they contain because files in either format include a header
that contains information such as the sample rate and the number of bytes of
data in the file.
88
Figure 4.4. Single file name tab.
Figure 4.4. The File Name tab of the Configure New Recorder dialog
box.
Directory You specify where to save acquired files either by typing a pathname (e.g.,
c:\SoundData\wrenProject\05May2002\) or by browsing through
your computer’s file system. Click the Browse... button to obtain a
browser dialog that lets you choose a directory (folder) in which to save
acquired files.
Sound File(s) You specify the name of the file(s) in the Sound File(s) field. For single-file
acquisition, the file name can be a simple text string that you type directly
into the Sound File field, or it can be a file name template that incorporates
date and time information that Raven fills in at the time acquisition starts,
as discussed below.
For file sequences, you must specify a file name template that Raven uses
to build a unique name for each file in the sequence. A file name template
contains some combination of date, time, and sequence number tags
(Table 4.1). When each file in a sequence is created, Raven substitutes the
89
corresponding value for each tag in the template to generate the file
names.
Table 4.1. Date, time, and sequence tags used in acquired file names.
Minimum and maximum values for date and time tags are shown in
brackets.
Tag
Definition
<yy>
year; 2 digits minimum
<ll>
month [1 12]; 2 digits minimum
<dd>
day [1 31]; 2 digits minimum
<hh>
hour [0 23]; decimals permitted;
2 digits minimum
<mm>
minute [0 59]; decimals permitted;
2 digits minimum
<ss>
second [0 59]; decimals permitted;
2 digits minimum
<n>
integer sequence number, starting at 1
The date and time values substituted for tags in the template depend on
the Start Time setting, as discussed under “Start Time” on page 91.
The number of characters that occur within a date, time, or sequence tag
determine the number of characters that are written to represent that unit
in the file name. For example, if the template is “set1-<nn>”, the resulting
filenames will end with two-digit sequence numbers; while “set1-<nnn>”
would result in filenames ending three-digit sequence numbers.
90
For hour, minute, second, and date tags, you can specify decimal values.
See Table 4.2 for examples.
Table 4.2. Examples of file names generated by various file name
templates, assuming a start time of 10:30:00, and a file size of 20 seconds.
Template
Resulting file names
BCV-<hh><mm><ss>.aif
BCV-103000.aif
BCV-103020.aif
BCV-103040.aif
BCV-103100.aif, ...
BCV-<hh><mm.mm>.aif
BCV-1030.00.aif
BCV-1030.33.aif
BCV-1030.67.aif
BCV-1031.00.aif, ...
BCV-<hh.hhhh>.aif
BCV-10.5000.aif
BCV-10.5056.aif
BCV-10.5111.aif
BCV-10.5167.aif, ...
BCV-<n>.aif
BCV-1.aif
BCV-2.aif
BCV-3.aif
BCV-4.aif, ...
BCV-<nnn>.aif
BCV-001.aif
BCV-002.aif
BCV-003.aif, ...
When you enter a filename containing date or time tags, the string labelled
“Example:” below the Audio File(s) field shows what the filename would
be once the relevant date and time values are substituted for the tags.
Start Time You can either specify that Raven use the date and time given by the
computer’s internal clock (“Now”), or you can specify some other date
and time (such as when the recording was actually made, if you are
acquiring from tape). Times are specified using a 24-hour representation; 1
PM would thus be represented as 13.
File name collisions When recording to a file or file sequence, if a file to be created has the
same name as a file that already exists in a given directory, Raven by
default creates a new subdirectory and puts the new file there, in order to
avoid overwriting the old file. The new subdirectory is named
“Recordings-date.time” where date and time identify the time when the
directory was created, using the format
<yyyy><ll><dd>.<hh><mm><ss> (see Table 4.1 for more details).
91
When this happens, Raven displays a message showing the name of the
file and the directory into which it was placed.
Alternatively, you can configure Raven to overwrite existing files when
file names collide by editing an entry in the Raven preference file, as
described in “Overwrite behavior for recorders” in Chapter 11 (page 309).
Recording to a file sequence
If you choose Record to File Sequence, the File Names tab appears
(Figure 4.5). The File Names tab contains the same fields as shown on the
File Name tab used for single-file recording (see “File naming” on
page 88), with the addition of a List File field.
Figure 4.5. File sequence names tab.
Figure 4.5. The File Names tab of the Configure New Recorder dialog
box.
When you record to a file sequence, Raven saves the incoming audio data
to a continuous series of files as described in “File Format” on page 86 and
“File naming” on page 88. In addition, Raven saves a list file, which is a
plain text file containing a list of all of the audio files recorded in the file
sequence. Once an audio file sequence and its associated list file have been
created, you can open the sequence as a single signal by opening the list
92
file. Opening a file sequence via a list file is discussed in “Opening file
sequences” in Chapter 7 (page 194).
If file name collisions (see “File name collisions” on page 91)
occur while acquiring to a file sequence, the list file and some
of the acquired audio files can end up in different directories.
Raven will warn you that this has occurred, and then automatically write the correct relative pathnames into the list file.
You can open the list without making any changes to it.
Note that if you configure the name of your list file with relative times in
the name of its sub-folder, Raven will generate a new list file whenever the
relative times you used are changed. For example, see the following list
file name templates:
raven.<yyyy><ll><dd>.<hh><mm><ss>.txt - this will result in a single list
file for all of your recordings, with the list file’s name determined by when
the first recording starts.
raven.<yyyy><ll><dd>/<hh><mm><ss>.txt - this will result in a new list file
being created in a separate folder for each new day that passes. If you
choose this option, you may wish to configure the names of your
recordings so that they will be separated into the same folders as their list
files, to allow you to analyze them in a more straightforward manner.
Using recorder presets
You can use recorder presets to save and retrieve sets of parameters for the
entire recorder dialog, and for each tab within the dialog. To save a
recorder preset for the entire dialog (including parameters on all tabs),
first configure all of the parameters on all of the tabs in the dialog the way
you want to save them, then choose Preset > Audio Recorder > Save As...
from the dialog’s menu bar. When the Save Audio Recorder Preset dialog
appears, enter a name for the preset, and click OK.
After making changes to an existing recorder, you can save those changes
as a recorder preset in the Configure Recorder dialog (View > Configure
Recorder...). If you have made changes to your window layout and you’d
like to preserve them as part of your recorder preset, you must first save
your changes as a new window preset (View > Window Preset > Save
93
As...), and then select that window preset on the Display tab of the
Configure Recorder dialog before you save the recorder preset.
Audio recorder presets must be saved in the folder Presets/Audio Recorder/ within the Raven program folder.
You can also create additional folders within the Audio
Recorder folder by clicking on the New Folder icon within
the Save Sound Window dialog. These folders will appear as
submenus in the Preset > Audio Recorder menu, with each
submenu listing the presets in the corresponding folder.
To save a preset for a single tab, configure the parameters on that tab, then
choose Preset > tab-name > Save As... from the dialog’s menu bar. For
example, to save a preset for the Input tab (while showing the Input tab)
choose Preset > Audio Recorder Input > Save As... . When the Save dialog
appears, enter a name for the preset, and click OK.
To retrieve a preset for the entire recorder dialog, choose Preset > Audio
Recorder > preset-name. To retrieve a preset for the current tab, choose
Preset > tab-name > preset-name. When you retrieve a preset, all of the
parameters in the Configure Recorder dialog or in the current tab are
immediately set to the saved values. If you then change some parameters
and want to revert to the saved values, choose the name of the preset from
the Preset menu again. If you want to save changes you’ve made under
the name of the last preset you loaded, choose Preset > Audio Recorder >
Save “preset-name” or Preset > tab-name > Save “preset-name”.
The recorder window: real-time views during recording
When you click OK in the Configure New Recorder dialog, a new recorder
window appears on the Raven desktop (Figure 4.6), with the window
layout that you specified on the Display tab. If you chose recording to
memory in the Configure New Recorder dialog, only the triangular green
Record-to-Memory button appears at the bottom of the window. If you
chose recording to a file or file sequence, the Record-to-Memory and
round red Record-to-Disk buttons both appear at the bottom of the
window.
Except for the two record buttons at the bottom of the window, a recorder
window looks and behaves exactly like any other Raven sound window. It
can contain multiple views that can be linked to each other in various
ways. By default, a new recorder window contains a single waveform
view. You can make selections in a recorder window and see
corresponding measurements in the selection table.
94
Figure 4.6. New recorder window
Figure 4.6. A new recorder window, configured for recording to file(s),
with a 30 second recording buffer.
Creating real-time You can create a new view in a recorder window any time, whether the
signal views recorder is running or stopped. If a recorder window already contains
multiple views, particularly if they are displayed at a large size, Raven’s
responsiveness to keyboard and mouse activity may slow down
noticeably on slower computers while the recorder is running. In such
cases, it may be preferable to create new spectrogram or spectrogram slice
views while the recorder is stopped.
Recording to To begin recording to memory, click the Record to Memory button at the
memory bottom of the window. Data begin appearing at the right edge of the
window (where the position marker is located by default in a recorder
window) and scroll across from right to left in all of the waveform and
spectrogram views that are displayed. If data do not start appearing in the
recorder window, refer to “Problems with recording” on page 107. To stop
recording, click the square green Stop Recording button that replaces the
Record to Memory button.
Each time you start recording, any data already in the recording buffer are erased.
95
Recording to files Raven always records to memory when recording to a file or file sequence,
so that the recorder window can display real-time views during
acquisition. To begin recording to a file or file sequence, click the Record
to File button. The Record to File button is then replaced by the square red
Stop-Recording to File button. If the recorder is not already recording to
memory, recording to memory and to a file or file sequence begins
immediately, and all views in the window begin displaying data. If the
recorder is already recording to memory when you click the Record to File
button, recording to memory continues, and recording to the specified file
or file sequence begins.
While recording to a file, Raven displays the name of the file and a
progress bar at the bottom of the recorder window to indicate how much
of the file has been recorded. If you’re recording to a single file, file
recording stops when the file has reached the size specified in the
Configure Recorder dialog (see “File Size” on page 88). If the recorder was
recording to memory before file recording started, it continues recording
to memory after file recording stops. When a file in a sequence has
reached the specified size, recording to disk continues uninterrupted with
the next file, and continues until you click the Stop Recording to File
button.
If you are recording from a tape player and would like to hear
some of the recorded sound just after it has been recorded,
you can use the Raven scrolling play button to do this. If you
use the play button, you will hear the current buffer from its
beginning, not the sound as it is being recorded. Also, if you
are using a microphone to record, you should not attempt to
play the sounds as they are being recorded because the output sounds from the speakers will feed back into the microphone. This issue may be avoided if headphones (as
opposed to speakers) are used to listen to the sounds.
Adjusting recording In order to avoid distortion in recorded signals, it is important to adjust
level the recording level appropriately. You adjust the recording level using
controls provided by your operating system, as discussed in Appendix C,
“Configuring Audio Input”.
Ideally, the level should be adjusted so that the minimum and maximum
excursions of the waveform view are close to the vertical limits of the
waveform scale, without actually reaching those limits. If the recording
level is set too low, some features of the signal may not be visible in signal
views. If the recording level is set too high, the acquired waveform will be
clipped. Clipping occurs when the magnitude of the original analog
waveform exceeds the maximum magnitude that can be represented with
the given sample size and recording level. The result is that waveforms get
truncated or “clipped” in the digital signal (Figure 4.7). When viewed in
the frequency domain (i.e., in spectrogram or spectrogram slice views),
96
clipped signals include spurious harmonics not present in the original
analog signal (see spectrograms in Figure 4.7).
Figure 4.7. Clipped and unclipped signals
Figure 4.7. (a) Unclipped and (b) clipped waveform and spectrogram
views of a 500 Hz sinusoidal signal. The clipped version was acquired
with the recording level set too high. The unclipped version accurately
represents the original signal. Note how the tops and bottoms of the
waveform are truncated in the clipped version, resulting in the
appearance of spurious harmonic energy at 1000, 1500, and 2500 Hz
in the clipped spectrogram.
Playing sound while While a recorder is running, you can choose to either listen to or mute the
recording sound as it is being recorded. Two buttons to the right of the play button
at the bottom of the recorder window provide these options.
97
Figure 4.8
Figure 4.8. The Listen/Mute while recording buttons are available at
the bottom left corner of the recorder window.
Navigating and Zooming and scrolling
selecting data in a When you first create a recorder window, the time axis is (by default)
running recorder zoomed so that the length of the recording buffer just fits in the window.
The time position location is at the right-hand edge of the window. As
with any other sound window, you can zoom and scroll any view in any
dimension, whether the recorder is running or stopped. If you drag the
scroll thumb away from the right-hand end of the scrollbar while the
recorder is running, the view stops scrolling and the data remain
stationary in the window, although acquisition continues without
interruption. When you release the scroll thumb at any point other than at
either end of the scrollbar, the thumb begins to “crawl” to the left. This
movement occurs because, as the view remains fixed at a particular time,
its position within the buffer changes as new data are acquired and the
oldest data are discarded. When the scroll thumb approaches the left edge
of the window (representing the point at which the oldest data are
discarded), data begin to disappear from the left-hand end of the view,
and are replaced from left to right by a gray background.
Selections and editing in a recorder window
When a recorder window contains audio data, you can create selections,
whether the recorder is running or stopped. You can copy and save
selections, but Cut, Delete, Paste, Filter and Amplify operations are
prohibited in recorder windows.
Reconfiguring an Reconfiguring while acquisition is stopped
existing recorder When a recorder is stopped, you can reconfigure all of its parameters.
Choose Configure Recorder... from the recorder window’s context menu
to open the Configure Recorder dialog box. You can change any of the
parameters in this dialog, and click OK.
When you reconfigure a recorder, any data that are already in
the recording buffer are lost, and any detectors in the
recorder are deleted.
98
Reconfiguring while recording
While a recorder is running, you can change its update rate and display
latency (see “Update Rate” on page 85, and “Latency” on page 85). Other
parameters cannot be changed without first stopping the recorder. Choose
Configure Recorder... from the recorder window’s context menu to open
the Display tab of the Configure Recorder dialog. In the dialog, there is an
asterisk on those tabs that can be modified while the recorder is running.
Multiple recorder Raven allows you to create multiple recorders, with different parameters.
windows If you need to switch back and forth repeatedly between different sets of
parameters— e.g., different sample rates or different destinations for
acquired files— you can keep multiple recorders open on the Raven
desktop, and switch among them as needed. However, only one recorder
at a time can be running.
Advanced options: acquiring from recordings at altered input speeds
Raven can correct for alterations in the speed at which a recorded sound is
played during acquisition.
It is sometimes necessary or useful to acquire sounds from a recording
that is being played back either slower or faster than the original recording
speed. For example, ultrasonic signals of bats and dolphins cannot be
properly digitized using most computer sound cards, because these cards
can only digitize sounds up to 22 kHz (half the maximum sample rate on
most sound cards) without aliasing.3 One solution is to shift the
frequencies in the recorded input sound down below 22 kHz during
acquisition by playing the sound back at a reduced speed.
Input speed adjustment is useful only when acquiring from a
recording that can be played back at altered speed. If you
need to record ultrasonic signals directly into Raven from a
live source, you need to have a sound card that provides a
sample rate at least twice the highest frequencies in the signal. For further information about sample rates, see “Sample
rate” in Appendix A (page 320).
When working with signals that are too low in frequency to be audible, it
may be convenient to play the tape back at a high speed during
acquisition, so that you can hear the signals as they are acquired.
If you click on the Advanced button at the bottom of the Configure
Recorder dialog, an additional tab labelled Speed becomes available
3. Aliasing is a type of distortion that occurs when sounds that are being digitized contain energy at frequencies that are more than half the sample rate.
Aliasing is explained more fully in “Aliasing and the Nyquist frequency” in
Appendix A (page 320).
99
(Figure 4.9). The Speed tab lets you specify whether the sound to be
acquired is at normal speed, or slowed down or sped up by a specified
factor relative to normal speed. To hide the Speed tab, click the Simple
button that appears in place of the Advanced button.
For example, to shift signals from 150 kHz down so that they could be
acquired by a sound card with a maximum bandwidth of 22 kHz, you
could play the recording at 1/8 normal speed during acquisition (150 kHz/
8 = 18.75 kHz). In this case, you would specify in the Speed tab that the
recording was slowed down by a factor of 8.
If you provide the appropriate input speed correction factor in the Speed
tab, then Raven will display time and frequency axes of views, and
measurements so that they refer to the original sound at normal speed.
Sounds that are acquired at an altered speed using the
appropriate speed correction factor behave in Raven as if
they had been acquired at normal speed. Thus, to hear
sounds that were originally above or below the human hearing range, you will need to specify an appropriately low or
high speed for playback, as described in “Playback rate” in
Figure 4.9. Speed tab
Figure 4.9. The Speed tab of the Configure New Recorder dialog box.
This tab is visible only after you click the Advanced button (which is
then replaced by the Simple button).
100
Advanced options: sample rate conversion during recording
Raven can decimate a signal as it is being acquired. Decimation is a way of
reducing the sample rate of a signal by retaining every Nth sample, and
discarding all others. Raven automatically applies a low-pass filter to a
signal before decimating it, in order to prevent aliasing in the decimated
signal.4
Why decimation is For low-frequency sounds, the lowest sample rate available with a
useful particular device may be many times higher than needed to represent the
signal accurately. This can lead to digitized signals that are much larger
than necessary, which require more storage and longer processing times.
Decimation during signal acquisition is a way of reducing the sample rate
to a rate that is more appropriate for the signal. Appendix A, “Digital
Representation of Sound” discusses choice of sample rate further.
For example, most of the sounds made by blue whales or elephants are at
frequencies below 100 Hz. For many types of analysis of these signals,
sample rates of 200 to 500 Hz are desirable. However, the lowest sample
rate provided by many recording devices is 8000 Hz. By choosing a
decimation factor of 16 or 32, the signal can be acquired with a sample rate
of 500 or 250 Hz, respectively.
Decimation Factor If you click on the Advanced button at the bottom of the Configure
Recorder dialog, an additional tab labelled Rate Conversion is displayed
(Figure 4.10). You can hide the Rate Conversion tab by clicking the Simple
button, which replaces the Advanced button.
4. Aliasing is a type of distortion that occurs when sounds that are being digitized contain energy at frequencies that are more than half the sample rate.
Aliasing is explained more fully in “Aliasing and the Nyquist frequency” in
Appendix A (page 320).
101
Figure 4.10. Decimation tab.
Figure 4.10. The Rate Conversion tab of the Configure New Recorder
dialog box. This tab is visible only after you click the Advanced button
(which is then replaced by the Simple button).
The Rate Conversion tab displays the Input Sample Rate, which is the rate
selected on the Input tab (see “Sample Rate” on page 83), a drop-down
menu from which you can choose a decimation factor, and the resulting
Output Sample Rate. The Output Sample Rate is the sample rate of the
signal that will be acquired. A decimation factor of N means that every
Nth sample is retained, resulting in a sample rate of 1/N times the input
rate. Thus, a decimation factor of 1 implies no decimation. The decimation
factor available are 2, 4, 8, 16, and 32.
Make note about decimation presence on input tab?
Advanced options: Amplified recording
When recording weak signals, you can opt to amplify the signal as it is
recorded, as opposed to recording the weak signal and applying
amplification later. To configure amplification for a new recorder, click on
the Advanced button at the bottom of the Configure Recorder dialog and
enter the desired amplification factor in the Amplify tab.
Advanced options: Adaptive filtering
You can also choose to apply an adaptive line enhancement filter to the
signal as it is recorded. This may help to reduce the amount of
102
background noise that is recorded with a signal. To configure line
enhancement, click on the Advanced button at the bottom of the Configure
Recorder dialog, and choose the Line Enhancer tab. To enable filtering,
select the Enable ALE checkbox, and then adjust the filter parameters as
desired. These parameters are the same as those for the Adaptivity filter,
which are described in the section “Adaptive Filtering” in Chapter 6 (page
148).
Advanced options: scheduled recording
If you click on the Advanced button at the bottom of the Configure
Recorder dialog, an additional tab labelled Schedule becomes available
(Figure 4.11). The Schedule tab contains parameters that you can set to
control when Raven begins acquiring data. To hide the Schedule tab, click
the Simple button that appears in place of the Advanced button.
Figure 4.11. Recording schedule tab.
Figure 4.11. The Schedule tab of the Configure New Recorder dialog
box. This tab is visible only after you click the Advanced button (which
is then replaced by the Simple button).
Retroactive The Retroactive Recording Offset field allows you to specify the time
Recording Offset offset used in retroactive recording to files. When a recorder is already
recording to memory, retroactive recording enables you to start recording
to a file or file sequence beginning at a time in the recording buffer before
the moment that you clicked the Record-to-Disk button. Retroactive
103
recording is discussed further in “Advanced options: retroactive recording
to files” on page 104.
Advanced options: retroactive recording to files
When configuring a recorder, you can use the Retroactive Recording
Offset field to specify that acquisition to disk files should begin at a point
in time prior to the moment when you click the Record to File button, provided
that the recorder was already recording to memory (see “Retroactive
Recording Offset” on page 103). This feature is useful when acquiring data
from field or lab recordings that contain signals of interest interspersed
among long segments without such signals. For example, if you’re
acquiring data from a tape recording, you might configure a file sequence
recorder with a 5 second retroactive recording offset. You would then
monitor the tape by starting the recorder in record-to-memory mode and
watching real-time views as the tape is played back. When you see or hear
a signal of interest, you would click the Record to File button. Raven
would immediately begin acquiring data, starting with the last 5 seconds in
the recording buffer (i.e., the last five seconds shown in real-time views in
the recorder window). As long as you start the disk recording within 5
seconds of the start of your signal of interest, that signal would be saved to
disk.
Advanced options: exporting clip files while recording
The Clip Exporter allows sound clips to be saved before the audio data in
the real-time recorder leaves the audio buffer. This facilitates making
manual selections in a recording while it is recording. When a selection is
committed, the audio data in the selection is saved to a file.
Choosing “Export selections to clip files” displays the Clip Format and
Clip Names tabs, which allow you to specify how selections are saved. The
options in these tabs are analogous to those in the File Format and File
Names tabs with the exception of a few additional settings. (For
descriptions of the File Format and File Names tags, see “File Format” on
page 86 and “File naming” on page 88) In addition to options present in
the File Format tab, the Clip Format tab allows you to configure the pad
size and choose which channels to export with a selection. Choosing a non
zero pad size includes data up to the specified time before and after each
selection in the exported clip. The Clip Format and Clip Names tabs are
shown in Figure 4.12.
The Clip Exporter is an alternative to the Schedule tab (see “Advanced
options: scheduled recording” on page 103) with the Record To File
option. The Schedule tab allows a recording to be made each time the red
record button is pressed. However, the system needs to be configured to
do retroactive recording in order to record sounds that are already in the
buffer, and the record button needs to be pressed before the end of the
104
retroactive time limit. In contrast, the Clip Exporter allows a selection to be
made at any time within the 30-second buffer (Figure 4.13).
Also, the Clip Exporter is similar to Save Active Selection As; however, it
allows all of the names that will be assigned to the files to be
preconfigured, since there isn’t time to fill in this information before the
sound leaves the buffer.
The Clip Exporter can also be used in conjunction with a
Raven detector (see “Saving detected selections” in Chapter
10 (page 273)) so that when an event signal is detected, a
selection is created, and the selection is automatically saved
to a file.
105
Figure 4.12 Clip tabs
Figure 4.12. (a) The Clip Format tab and (b) the Clip Names tab within
the recorder window.
106
Figure 4.13 Clip Exporter
Figure 4.13. A recorder window with the “Export selections to clip files”
option selected.
Problems with recording
Device unavailable If, when you click on the Start Recording button in a recorder window,
Raven displays a message indicating that the audio device is unavailable,
check to make sure that there is not another recorder running within
Raven. Only one recorder at a time can be running. Another possible
reason for this condition is that the audio input device has been allocated
by another application. If there is another audio application running, you
may need to quit from it in order to record in Raven. Some audio CD
player programs allocate the audio device for their exclusive use. Such
applications cannot be used to play CDs for input into Raven. If you have
this problem, try using a different program for playing CDs.
Unsupported audio While using some audio CD player programs, attempts to record with
format Raven may result in an error message stating that “the specified audio
format is not supported.” In most cases, this message is itself erroneous
(i.e., the audio format is supported), and the problem can be rectified by
using a different audio CD player program.
Frame rate monitor While a recorder is running, Raven measures the sample rate from the
recording device. If the measured recording rate is different than the
designated setting, a notification dialog will be displayed to alert the user
of this discrepancy. This may indicate that the recording device is not
digitizing properly, or that samples are being dropped in the recording
107
process. If this occurs, please contact the Raven team via the email
feedback tool.
If your recording device regularly results in this message
being displayed and you are confident it is recording as
intended, you can turn off the frame rate monitor or change its
threshold by modifying the corresponding preferences entry.
To turn the frame rate monitor off, change the entry
raven.ui.audioRecorder.sampleRateMonitor.enable from true
to false.
108
Chapter 5
Spectrographic Analysis
About this chapter
This chapter is a reference for working with spectrogram, spectrogram
slice, and selection spectrum views in Raven. The chapter assumes that
you have read Chapter 1, “Getting Started”, and that you are acquainted
with basic concepts involved in short-time Fourier transform (STFT)
analysis of time-varying signals. Terms such as data record (equivalent to
“frame” or “aperture”), filter bandwidth, window overlap, and window
function are explained here only briefly. If you are not already familiar
with these concepts, we recommend that you read Appendix B, “A
Biologist’s Introduction to Spectrum Analysis”, which provides the
conceptual background needed to make full use of Raven’s spectrogram
analysis capabilities.
About spectrographic views
Raven provides three types of views that show the relative intensity of
different frequency components of a sound (i.e., sound spectrum
information; see Figure 5.1).
•Spectrogram views show how the spectrum of a sound varies over time
in a three-dimensional plot, in which time is represented on the
horizontal axis, frequency on the vertical axis, and the relative power at a
given point in time and frequency as a color (by default, grayscale)
value.
•Spectrogram slice views display a single spectrum from the series of
spectra that constitute a spectrogram (see below), plotted as a twodimensional line graph with frequency on the horizontal axis, and
relative power (in decibels) on the vertical axis.
•Selection spectrum views show the average spectrum of a sound over
the time interval defined by the active selection.
109
Chapter 5: Spectrographic Analysis
Figure 5.1. Three spectral views
Figure 5.1. Raven’s three spectrographic views. The spectrogram slice
shows a single spectrum from the spectrogram, at the time indicated by
the time position marker (8.329 seconds). The selection spectrum
shows the average spectrum over the time interval defined by the
highlighted selection in the spectrogram.
How the To calculate any one of the three spectrographic view types, Raven divides
spectrographic a sound (or a selected portion of a sound) into a series of successive short
views are related time segments, or records, and calculates a single spectrum for each record,
using an algorithm called the discrete Fourier transform (DFT). For
spectrogram and spectrogram slice views, this process is performed for
the entire sound. For selection spectrum views, only the part of the sound
within the time interval of the active selection is processed.
In a spectrogram, Raven displays the spectra of all of the records, arranged
successively from left to right according to the time of the corresponding
record in the sound. Each spectrum is displayed as a tall, narrow bar
(Figure 5.2, upper view). The vertical axis of each bar represents the
frequency axis of the spectrum. The color of the bar varies from bottom to
top, with the color at a particular frequency (i.e, height along the bar)
representing relative power or intensity at that frequency for the
corresponding record.
In a spectrogram slice, Raven displays only one spectrum at a time (Figure
5.2, lower view). The choice of which spectrum is displayed at any
moment depends on the time position of the spectrogram slice view, as
110
described later in this chapter (see “Time position of a spectrogram slice
view” on page 137).
Figure 5.2 Boxy gram and slice
Figure 5.2. Relationship between spectrogram and spectrogram slice
views. The spectrogram shown here is a close-up of about 1 second
from Figure 5.1, encompassing 17 individual spectra. Spectrogram
smoothing has been turned off to reveal the individual spectra in the
spectrogram. The highlighted selection identifies the one spectrum that
appears in the spectrogram slice view. The time position of the
spectrogram slice view is indicated by the time position marker in the
spectrogram (at 20.555 sec).
111
In a selection spectrum view, Raven displays an average spectrum based
on a series of short spectra computed over the time interval of the active
selection (Figure 5.1, bottom view).
Raven now introduces a slight change in the calculation of
power spectra. The new method includes energy in the negative frequency bins of the DFT, as well as normalization of the
power spectrum by the number of points in the DFT window.
Power measurements made using this method will differ from
the same measurements performed with previous versions of
Raven by a multiplicative factor of 2/Nfft (1/Nfft for the zero
frequency component). To use this updated power spectra
calculation, check the box on the option screen that appears
when Raven starts for the first time.
The new power spectra calculation should be used in cases
when measurements need to be consistent with those made
in Canary.
For more specific information on this new measurement
method, including how to change the current method and run
comparisons between the methods, see “Power Spectra Calculation Method” in Chapter 11 (page 304).
Since all three spectrographic views start with calculating spectra of a
series of records, most of the parameters that must be specified for all
three view types are the same. The next section briefly explains each
parameter.
Configuring spectrographic views
To create a new spectrogram, spectrogram slice, or selection spectrum
view, click on the appropriate New View button in the view toolbar
(Figure 5.3) or choose a view type from the View > New menu.
Figure 5.3. New View buttons
Figure 5.3. The New View buttons, in Raven’s view toolbar.
112
A dialog box appears, containing parameters for configuring the
requested type of spectrographic view (Figure 5.4). The dialog boxes for
configuring spectrogram and spectrogram slice views are identical, except
for their titles. The dialog boxes are identical because both view types
calculate a spectrogram of the entire sound; the only difference between
spectrogram and spectrogram slice views is in how the data are displayed
(see “How the spectrographic views are related” on page 110). The dialog
box for configuring a selection spectrum view is the same, except that it
lacks the Averaging parameter.
The remainder of this section explains each of the parameters in the
configuration dialog box.
Figure 5.4 Configure Spectrogram dialog
Figure 5.4. The Configure New Spectrogram dialog box.
Window type Each data record is multiplied by a window function before its spectrum is
calculated. Window functions are used to reduce the magnitude of
spurious “sidelobe” energy that appears at frequencies flanking each
analysis frequency in a spectrum. These sidelobes appear as a result of
113
analyzing a finite (truncated) portion of a signal. A window function can
reduce these sidelobes by “tapering” the portion of the waveform that
appears in each window. Window functions are discussed further in
Appendix B, “A Biologist’s Introduction to Spectrum Analysis”.
Raven provides six different window functions: Blackman, Hamming,
Hann (sometimes called Hanning), Kaiser, rectangular, and triangular
(sometimes called Bartlett). Each window function is characterized by the
magnitude of the sidelobes relative to the center lobe. The difference in
decibels between the center lobe magnitude and the magnitude of the
largest sidelobe is called the sidelobe rejection (Figure B.10 on page 342). In a
grayscale spectrogram, differences among windows in sidelobe rejection
result in different amounts of gray “fringing” above and below black or
very dark areas.
For a given window size, different window functions will result in
different filter bandwidths (see “3 dB Bandwidth” on page 117). In terms
of a spectrogram, this means that the vertical thickness of a horizontal line
representing a pure tone will depend on which window function is used.
Figure 5.5 and Figure 5.6 illustrate the effect of different window functions
on spectrogram and spectrogram slice views of the same signal.
114
Figure 5.5. Window comparison-
SPG
Figure 5.5. Effect of choice of window function on spectrograms. The
signal is a series of calls from a red-breasted nuthatch, digitized at 44.1
kHz. All three spectrograms have the same window size (= 512 points,
11.6 mS), hop size = 5.8 mS (frame overlap = 50%), and frequency grid
spacing = 86.1 Hz (DFT size = 512 samples). 3 dB bandwidths: (a) 141
Hz, (b) 124 Hz, (c) 76.2 Hz.
115
Figure 5.6. Window compare-
slice
Figure 5.6. Effect of window function on spectrogram slice views.
These spectrogram slice views were made at the point indicated by the
position marker in Figure 5.5, midway through the second call. All three
spectra have the same window size (= 512 points, 11.6 mS), and
frequency grid spacing = 86.1 Hz (DFT size = 512 samples). 3 dB
bandwidths: (a) 141 Hz, (b) 124 Hz, (c) 76.2 Hz.
The appearance of sidelobes in spectra of finite-length signals, the use of
window functions to reduce their magnitude, and differences among the
various window functions are discussed further in Appendix B, “A
Biologist’s Introduction to Spectrum Analysis”.
Window size The Window Size parameter controls the length of each data record that is
analyzed to create each of the individual spectra that together constitute
the spectrogram. You can specify window size either in number of
samples from the digitized signal, or in time units (seconds or
milliseconds) by choosing the preferred unit from the drop-down menu.
The default unit is samples. If you specify window size in seconds or
milliseconds, Raven uses the number of samples that most closely
approximates the window size that you enter.
The maximum value of the Window Size parameter depends on whether
the DFT Size parameter is locked, as discussed in “Frequency grid spacing
and DFT size” on page 122. When DFT Size is unlocked, Window Size can
116
be set to a maximum of 65,536 samples (= 216). When DFT Size is locked,
the maximum value of the Window Size is equal to the DFT Size.
Window Size slider control
Adjacent to the Window Size field is a slider control that provides an
alternate means for changing the window size. Sliding the control to the
right increases the window size. The control is logarithmic: the farther the
slider is moved to the right, the more the window size changes in response
to a given movement. The window size slider is useful primarily when the
Auto-apply checkbox is checked (see “Apply and Auto-apply” on
page 129). When Auto-apply is turned on, Raven recalculates the
spectrogram immediately as you adjust the slider, allowing you to
instantly see how changes in window size affect the tradeoff between time
and frequency resolution.
Use of the Window Size slider with Auto-apply turned on may
result in unacceptable delays in redrawing the spectrogram
with longer signals and/or slower computers.
Beta (Kaiser For the Kaiser window, you can set an additional parameter, called Beta,
window only) to values between 0 and 20. For a given window size, higher values of
Beta result in larger filter bandwidths and smaller sidelobes.
3 dB Bandwidth 3 dB bandwidth is the filter bandwidth of the individual analysis filters in
the filterbank simulated by the short-time Fourier transform (STFT) with
the selected window type and size (see Appendix B, “A Biologist’s
Introduction to Spectrum Analysis”). Specifically, the 3 dB Bandwidth
field displays the width (in Hz) of the main lobe of the spectrum of a
sinusoid at the point where the power is 3 dB lower than the maximum
power in the spectrum (Figure 5.7).
Figure 5.7. 3dB bandwidth
117
Figure 5.7. Spectrum of a pure tone sinusoidal signal. The 3 dB
bandwidth is the width BW (in Hz) of the spectrum’s main lobe at the
point where the power is 3 dB less than the maximum power in the
spectrum.
When you change the window size or the window type, the 3 dB
Bandwidth field is immediately updated to display the corresponding
bandwidth. For a given window type, improved time resolution (shorter
windows) inevitably results in poorer frequency resolution (larger
bandwidths). You can edit the 3 dB Bandwidth field to specify a desired
value directly. When you press <Enter>, click on another field in the
dialog, or click OK or Apply, Raven will choose the window size that
results in the closest available approximation to the 3 dB Bandwidth value
you entered. For further discussion of the tradeoff between time and
frequency resolution in spectrograms, see Appendix B, “A Biologist’s
Introduction to Spectrum Analysis”.
Choosing the In a spectrogram, where you are typically interested in frequency
window size variations with time, the “best” choice of window size depends in part on
the nature of the signal, and on what features you are most interested in
observing or measuring. If you are most concerned with precise frequency
measurements, you will probably want to choose a large window size
(hence better frequency and poorer time resolution). If you want better
time resolution, choose a shorter window size; the bandwidth will then be
larger (poorer frequency resolution; Figure 5.8).
118
Figure 5.8. Cassin’s KB, win size compare
Figure 5.8. Effect of choice of window size on time and frequency
smearing in spectrogram views of Cassin’s kingbird sound (digitized at
44.1 kHz). For both views, window type = Hann, hop size = 64 samples,
Frequency grid spacing = 22 Hz. (a) Window size = 800 samples, 3 dB
bandwidth = 79 Hz. (b) Window size = 150 samples, 3 dB bandwidth =
423 Hz. View (a) has better frequency resolution (note sharpness of the
nearly constant-frequency bands in selection #2), but poorer time
resolution (note horizontal smearing of the rapid downsweep in
selection #1, and in the oscillating frequencies in the second part of the
call).
Figure 5.9 shows an extreme example of how choice of window size can
change the appearance of a spectrogram. See Appendix B, “A Biologist’s
Introduction to Spectrum Analysis” for further discussion and more
examples of the effect of varying window size and bandwidth.
119
Figure 5.9. Window size spg demo (dolphin)
Figure 5.9. Effect of varying analysis resolution on spectrograms. The
signal is part of a rapid series of clicks produced by a spotted dolphin,
digitized at 48 kHz. The period between clicks is about 1.4 mS,
corresponding to a frequency of about 720 Hz (= 1/0.0014). The two
spectrograms differ only in window size, and hence bandwidth. In both
spectrograms, hop size = .208 mS, window = Hamming. (a) Bandwidth
= 3121 Hz (window size = 20 points = .417 mS), overlap = 50%. In this
representation, each click appears as a broad-band vertical stripe on
the spectrogram because the window size is short enough to resolve
individual clicks. (b) Waveform. When played at normal speed, the
signal sounds to a human like a buzz. (c) Bandwidth = 61 Hz (window
size = 1024 points = 21.3 mS), overlap = 99%. In this representation,
individual clicks cannot be resolved because each window
encompasses about 15 clicks; instead the click repetition frequency
appears as a series of horizontal bands spaced 720 Hz apart (the click
repetition frequency).
Time grid: Window Hop size1 refers to the time interval (measured either in samples or in time
Overlap and units such as seconds or milliseconds) between the beginnings of
Hop Size successive windows or records. In an unsmoothed spectrogram (see
“Smoothed vs. unsmoothed display” on page 134), the hop size can be
seen as the width or duration of the individual cells in the spectrogram
(Figure 5.10). Hop size can be smaller than the window size because
successive windows can overlap each other. Windows can also be
contiguous (0% overlap) or separated by time intervals that are omitted
from the analysis (negative overlap).
1. Hop size was called time grid spacing in versions prior to Raven 1.2.
120
Window overlap is usually expressed as percent of window size. For
example, an overlap of 50% means that each window begins halfway
through the preceding window. An overlap of -100% means that one
window of data is skipped between successive windows that are
analyzed; -300% skips three frames, and so on. The relationship between
hop size and window overlap is given by
hop size = window size * (100% - overlap%).
The Hop Size and Overlap fields in the dialog box are coupled so that you
can specify hop size either directly, by typing a value in the Hop Size field,
or indirectly, by typing a value in the Overlap field. Using the units dropdown menu, you can specify the measurement units for hop size as either
samples (the default), seconds, or milliseconds. If you enter a value in the
Overlap field that does not correspond to an integer number of samples,
Raven substitutes the closest overlap value that does.
Figure 5.10 shows three spectrograms that differ only in hop size.
121
Figure 5.10. Window overlap comparison
Figure 5.10. Effect of varying hop size in spectrograms. The signal is
part of a song of a lark sparrow, digitized at 44.1 kHz. The three
spectrograms are unsmoothed and differ only in hop size (window
overlap). In all three spectrograms, window type = Hann, window size =
512 samples (= 11.6 mS; 3 dB bandwidth = 124 Hz), frequency grid
spacing = 86.1 Hz (DFT size = 512 samples). (a) Hop size = 11.6 mS
(window overlap = 0%). (b) Hop size = 5.8 mS (window overlap = 50%).
(c) Hop size = 1.1 mS (window overlap = 90%).
A spectrogram made with a negative window overlap ignores
some of the available data, and can give an extremely misleading picture of a signal. Negative window overlaps should
generally be avoided unless you have some specific reason
for wanting to omit some parts of a signal from analysis.
Lock Overlap vs. Lock Hop Size
Next to the Overlap and Hop Size fields are two buttons, marked with
open and closed padlock icons. The button that displays the closed
padlock indicates which value— window overlap or hop size— will be
locked or held constant when you make changes to the window size.
Clicking on either button reverses the state of both buttons.
Frequency grid The frequency grid spacing of a spectrogram (visible as the height of the
spacing and DFT individual boxes in an unsmoothed spectrogram; see “Smoothed vs.
size unsmoothed display” on page 134) depends on the sample rate (which is
122
fixed for a given digitized signal) and a parameter of the STFT called DFT
size.2 The relationship is
frequency grid spacing = (sampling frequency) / DFT size
where frequency grid spacing and sampling frequency are measured in
Hz and DFT size is measured in samples. DFT size is constrained to be a
power of 2 that is greater than the current window size.
The DFT Size and frequency Grid Spacing fields in the Configure
Spectrogram dialog box are linked: you can specify the frequency grid
spacing either directly by choosing a value from the Grid Spacing dropdown menu, or you can choose a value from the DFT Size drop-down
menu. The DFT Size menu displays powers of 2 greater than or equal to
the current window size. Larger DFT sizes correspond to smaller
frequency grid spacings.
Lock DFT Size
Next to the DFT Size drop-down menu is a button marked with a padlock
icon. When this button is unlocked (the default), Raven adjusts the DFT
size as you change the window size, in order to maintain a consistent
relationship to the window size, subject to the constraint of being a power
of 2. For example, in the default spectrogram parameters, the DFT size is
the smallest power of 2 greater than or equal to the window size. If you
increase the window size from 512 to 513, Raven changes the DFT size
from 512 to 1024. If you manually choose the DFT size to be, for example,
the second power of two greater than or equal to the window size (e.g.,
with window size of 512, you set DFT Size to 1024 instead of 512), then
Raven will change DFT Size to maintain this relationship as you adjust
window size.
When the DFT Size padlock button is locked, the DFT size (hence
frequency grid resolution) is fixed, and will not change when the window
size changes.
Because the window size cannot exceed the DFT Size, the
maximum value you can specify for Window Size, either by
typing a value, or by moving the slider control, is limited to the
DFT Size value when DFT Size is locked.
Clipping level The Clipping Level parameter allows you to specify a “noise floor” below
which any amplitude value is altered, in order to reduce or eliminate the
2. The parameter that Raven calls DFT Size is sometimes called FFT size in other
programs. FFT stands for fast Fourier transform, which is a particular algorithm
used to compute the discrete Fourier transform or DFT. Size is a characteristic of
a particular DFT, not of the FFT algorithm used to compute it.
123
effect of the noise. After you enable clipping, you can modify it using two
parameters: (1) the power level below which values will be altered, and (2)
the power level at which to set the altered values.
Figure 5.11
Figure 5.11. The Clipping dialog box.
In a spectrogram view, if you choose to clip to a value of -Infinity,
everything that is below your noise floor will appear as white. If you clip
to a different value (for example, 0 dB), everything below your noise floor
will appear as very light. These differences will result in different
appearances once your spectrogram is smoothed, so you may wish to vary
your parameters and examine the results as you explore. See Figure 5.12
for an example.
In a spectrogram slice view, if you choose to clip to a particular dB value,
then the result will be a smooth slice view without any dropouts; but if
you choose to clip to -Infinity dB, the result will contain many dropouts
and will be more difficult to read. See Figure 5.13 for an example.
There are several reasons to perform clipping. First, because of the finite
precision of the digitization process, a digitized sound always contains
some error and has a limited dynamic range. For signals digitized with 8bit samples, the dynamic range is limited to 48 dB; for 16-bit samples, the
dynamic range is limited to 96 dB.3 Therefore any power value in an 8-bit
spectrogram that is more than 48 dB below the highest peak in the signal
must be noise introduced by the digitizing process, and should be
disregarded.4 The noise floor can also be useful for removing noise that
was present before the digitizing process (for example, from a recording
with low-level wind or other broad-band noise).
A more pragmatic reason for noise clipping is that very small power
values show up on a log scale as large negative dB levels (because the
3. The dynamic range of a digitized sound is 6 dB/bit.
4. If the highest spectral peak in a signal is smaller than the digitizer’s maximum
output level, the dynamic range between the peak and noise introduced by
digitizing will be less than 6 dB/bit. However, it can never be more than 6 dB/
bit.
124
logarithm of zero is negative infinity). The noise floor allows Raven to
ignore very small power values.
Finally, the noise floor can also be used to eliminate spectral sidelobes
(which show up as gray fringes around strong signal components in
spectrograms).
If the noise floor is set too low, excessive noise will be displayed in the
spectrogram or spectrogram slice along with the signal. If it is set too high,
portions of the signal will not be visible. You may need to experiment with
different clipping levels in order to find a value that produces a
satisfactory display.
As an alternative to clipping, you can alter the look of a spectrogram
without changing the underlying spectrogram data by altering the
brightness and contrast of the view. Note, though, that this method does
not change the noise floor in the spectrogram data stored in memory. If
you are only using spectrograms for visual examination and display, then
the distinction between brightness/contrast and the noise floor of the
spectrogram data is unimportant. However, if you plan to do any
quantitative analysis (e.g., correlations) using the spectrogram data,
remember that the only way to change the noise floor is to recalculate the
spectrogram, specifying a different Clipping Level in the spectrogram
dialog box.
If you want to raise or lower the noise floor in a spectrogram (either in the
display or in the data itself), you must recalculate the spectrogram with a
higher or lower Clipping Level.
Figure 5.12 and Figure 5.13 show spectrograms and spectrogram slices
that differ in Clipping Level as well as in brightness and contrast. Note
that if you want to have different clipping levels for your spectrogram and
spectrogram slice views, you must unlink those views in their
spectrogram parameters.
125
Figure 5.12
Figure 5.12. Effect of clipping level on spectrograms. The signal is part
of a song of a Black-capped Vireo. First row: no clipping level set,
standard brightness (50) and contrast (50) settings, spectrogram
smoothing enabled. Second row: no clipping level set, brightness set to
65, contrast set to 78, smoothing enabled. The underlying power values
have not changed so all measurement values will be the same as those
measured in the first row. Third row: clipping enabled, values below 60
dB clipped to -Infinity dB, spectrogram smoothing enabled.
Measurement values will differ between this view and the first row.
Fourth row: clipping enabled, values below 60 dB clipped to -Infinity dB,
spectrogram smoothing disabled. Without smoothing, more of the
clipped spectrogram is visible. Fifth row: clipping enabled, values below
60 dB clipped to 0 dB, spectrogram smoothing enabled. Since values
were only clipped to 0 dB, more of the clipped values are visible than in
the third row.
126
Figure 5.13
Figure 5.13. Effect of clipping level on spectrogram slices. All three
slices are on a single 256-point frame approximately 1.378 seconds
into the Black-capped Vireo song shown in Figure 5.12. The signal was
digitized with 16-bit resolution, and thus has a dynamic range of 96 dB.
The highest power values are near 90 dB, so there is little perceptible
noise inserted by the digitization process. First row: no clipping level
set. Second row: clipping level = 60 dB, with values clipped to 60 dB.
Third row: clipping level = 60 dB, with values clipped to -Infinity dB.
Notice the dropouts in the slice view. For this reason, we recommend
against clipping spectrogram slice views to -Infinity dB, but for analysis
purposes, you may want to clip spectrograms to -Infinity dB.
Spectrum The Averaging field allows you to specify the number of individual
averaging spectra over which Raven should average the power values to obtain the
values in each cell of the spectrogram. In most situations, Averaging
should be left at its default value of one spectrum.
Higher Averaging values may provide more satisfactory spectrogram
images when more than a few seconds of a signal are displayed. There are
two reasons why averaged spectrograms may be preferable at certain time
scales. First, if the time scale of a spectrogram view is such that the
number of spectra in the visible time span is much greater than the
number of pixels in the time dimension of the sound window, then many
spectra will not be displayed at all. Some acoustic events that span only a
few spectra may not be visible (unless you zoom in to display a finer time
scale) because the only spectra in which they appear fall between the
127
pixels shown in the display. By setting the Averaging field to a value
greater than one spectrum, you can make visible short-duration events
that would otherwise be lost between pixels. Second, spectrum averaging
smooths background noise, which can result in a higher signal-to-noise
ratio in the spectrogram image (Figure 5.14). At finer time scales (i.e.,
greater magnification in the time dimension), however, spectrum
averaging tends to blur signals (Figure 5.14).
You can specify the amount of data to average in units of seconds or
milliseconds (rather than spectra), using the units drop-down menu.
Figure 5.14. Spectrum averaging example.
Figure 5.14. Spectrum averaging can yield clearer spectrograms when
the interval in view is long compared to the number of spectra in view.
All four spectrograms of a nearby common yellowthroat and a distant
yellow warbler were made with Window Type = Hann, Window Size =
512 samples, Window Overlap = 50%. All four are linked by time
position. (a) and (c) Averaging = 1 spectrum. (b) and (d) Averaging = 4
spectra. When viewing a longer time span (views (a) and (b)), the view
that uses spectrum averaging provides a clearer image, especially of
the faint signal from a distant bird. For the more magnified image ((c)
and (d)), the view without averaging appears clearer.
128
Apply and Auto- If you click the Apply button, Raven immediately calculates and displays
apply the spectrogram, using the parameters currently displayed in the dialog,
without closing the dialog. (Clicking OK closes the dialog before
calculating the spectrogram.)
If the Auto-apply checkbox is checked, Raven immediately recalculates
and displays the spectrogram each time you change any parameter in the
dialog, without you needing to click the Apply button. For parameters that
you enter by typing in a field (e.g., Time grid spacing or window Overlap),
the spectrogram is recalculated when you complete an entry by pressing
the <Enter> or <Tab> key, or by clicking another field or control in the
dialog.
Spectrogram You can save and retrieve sets of spectrogram parameters using
presets commands on the Preset menu within the Configure Spectrogram dialog.
A set of saved spectrogram parameters is called a spectrogram preset. To
save a preset, choose Presets > Save As... . When the Save Spectrogram
Parameters dialog appears, enter a name for the preset, and click OK.
Spectrogram presets must be saved in the folder Presets/
Spectrogram Parameters/ within the Raven program
folder. You can also create additional folders within the Spectrogram Parameters folder by clicking on the New Folder icon
within the Save dialog. These folders will appear as submenus in the Preset menu, with each submenu listing the
presets in the corresponding folder.
To retrieve a spectrogram preset, choose the name of the preset from the
Preset menu. When you retrieve a preset, all of the spectrogram
parameters in the Configure Spectrogram dialog are immediately set to
the saved values. If you then change some parameters and want to revert
to the saved values, click the Reset button or select the name of the preset
from the Preset menu again. If you want to save changes you’ve made
under the name of the last preset you loaded, choose Preset > Save
“PresetName”.
Spectrogram views
Significance of the Spectrograms displayed by Raven have a logarithmic power (color) axis.
color (grayscale) That is, the color (by default, grayscale) values shown in the cells of an
values unsmoothed spectrogram represent the logarithm of the power at the
corresponding frequency for each spectrum in a spectrogram. Hence, the
color value is proportional to the power expressed in decibels (relative to
an arbitrary reference power).
The numeric values for relative power level associated with each point are
displayed in decibels (dB) in the mouse measurement field at the bottom
129
of the signal window. These decibel measurements are relative to an
arbitrary spectrogram bin reference value of 1.
Raven 1.4 does not display calibrated sound amplitude measurements. In other words, the dB values given by Raven do
not represent absolute sound levels. Raven’s amplitude measurements can be used to obtain accurate relative measurements within signals.
Brightness and The section on “Spectrogram brightness and contrast” in Chapter 3 (page
contrast 68) provides a basic introduction to the operation of the spectrogram
brightness and contrast controls. This section provides a more detailed
explanation of how these controls work. The information in this section is
not needed to use the controls effectively; it is provided for those who are
interested in a quantitative explanation of how these controls affect the
spectrogram image.
Each spectrogram has an associated floor value and ceiling value. Powers
that are less than or equal to the floor value are displayed as the floor color;
powers that are greater than or equal to the ceiling value are displayed as
the ceiling color. In a default grayscale spectrogram, the floor and ceiling
colors are white and black respectively. For power values between the
floor and ceiling values, Raven interpolates between the floor and ceiling
colors. Changing the brightness of a spectrogram changes the floor and
ceiling values by the same amount in the same direction. If we plot color
value as a function of power, changing the brightness has the effect of
shifting the brightness-vs.-power function left or right, without changing
its slope (Figure 5.15). Increasing the contrast shifts the floor and ceiling
values toward each other, which increases the slope of the brightness-vs.power function, without shifting the center point of the function (Figure
5.16).
130
Figure 5.15. Brightness graph
Figure 5.15. Relationship between color and power values in a default
grayscale spectrogram for two different brightness settings. f1 and f2:
power floor values for Brightness 1 and 2 respectively. c1 and c2: power
ceiling values for Brightness 1 and 2 respectively. g1 and g2: color
values associated with power P for Brightness 1 and 2 respectively.
Increasing the brightness (arrow) increases the power floor and ceiling
values by the same amount. The result is that the color associated with
a given power P gets lighter.
131
Figure 5.16. Contrast graph
Figure 5.16. Relationship between color and power values in a default
grayscale spectrogram for two different contrast settings. f1, f2, c1, and
c2 defined as in Figure 5.15. Increasing the contrast (arrows) decreases
the range of power values between white (floor) and black (ceiling).
When Raven initially displays a spectrogram, the floor and ceiling values
are set to values that yield acceptable spectrogram displays with a wide
range of signal amplitudes. The initial contrast and brightness settings are
arbitrarily assigned values of 50%.
In Raven 1.2, this provided acceptable spectrogram displays,
but if you choose to use the Raven 1.3 power spectrum computation, you may find that a brightness value of 33% might
be more appropriate because the power levels are lower.
If you set contrast to 100%, the floor and ceiling values are set equal to
each other, so that the spectrogram displays only two colors, which are the
floor and ceiling colors (by default, white and black). In this case,
adjusting the brightness moves the threshold between these two colors.
Time alignment of Each individual spectrum in a spectrogram has a time associated with it,
spectrogram data called the spectrum time. The spectrum time is the time halfway through
the block of samples from which the spectrum was calculated (the
spectrum source data). In an unsmoothed spectrogram, Raven displays each
individual spectrum in a spectrogram so that its horizontal midpoint is at
the spectrum time. Thus, in a waveform and spectrogram linked to each
other by time position and scale, the midpoint of each individual
spectrum is aligned with the midpoint of the corresponding source data
(Figure 5.17, Figure 5.18).
132
Figure 5.17.SPG align boxy 0% overlap
Figure 5.17. Time alignment of individual spectra in a spectrogram
made with window size = 10 samples, and time grid resolution = 10
samples (overlap = 0%). The waveform and spectrogram are linked in
time position and scale, and have been zoomed in to show individual
samples. The colored overlays identify the correspondence between
spectra and the samples from which they were calculated. Each
spectrum is associated with the time at the midpoint of its
corresponding data interval (dashed vertical lines). Compare with
Figure 5.18.
If the window overlap is set to 0%, the left and right edges of each
spectrum are aligned with the start and end of the source data (Figure
5.17). However, if the window overlap is greater than 0%, the edges of
individual spectra can’t be aligned with the edges of the sample window,
because the width of the displayed spectrum (i.e., the time grid resolution,
or hop size) is smaller than the width of the source data (Figure 5.18).
Whenever a spectrogram’s time grid resolution is less than the size of the
spectrum window (i.e., whenever window overlap is greater than 0%),
there is a gap between the time of the first sample in the signal, and the left
edge of the first spectrum in a spectrogram, where Raven displays the “no
data” color (gray by default; Figure 5.18). This gap is often narrow enough
to be unnoticeable or invisible unless you zoom in to a high magnification.
133
Figure 5.18. SPG align boxy 30% overlap
Figure 5.18. Time alignment of individual spectra in a spectrogram
made with window size = 10 samples, and time grid resolution = 7
samples (overlap = 30%). The colored overlays identify the
correspondence between spectra and the samples from which they
were calculated. Each spectrum is associated with the time at the
midpoint of the corresponding source data (dashed vertical lines). Note
the gray area of no data prior to the beginning of the first spectrum.
Compare with Figure 5.17.
Smoothed vs. Raven can display spectrograms in either of two modes: smoothed (the
unsmoothed default) or unsmoothed (Figure 5.19).
display
134
Figure 5.19. Smoothed and unsmoothed spg.
Figure 5.19. Close-up view of portion of two spectrogram views of a
signal digitized at a sample rate of 44.1 kHz. The upper view is
smoothed (the default); the lower view is unsmoothed. For both views,
window type = Hann, window size = 512 points, overlap = 50%.
To switch between smoothed and unsmoothed displays, click the
checkbox next to Smooth Spectrogram in the view’s contextual menu or in
the View menu when the spectrogram view is active.
An unsmoothed view more accurately represents the actual data in the
spectrogram. A smoothed spectrogram may provide a more visually
satisfying image. In an unsmoothed spectrogram, each actual data point
on the spectrogram grid is represented by a rectangular gray box. The
width and height of the boxes depend on the hop size and frequency grid
spacing respectively (see “Time grid: Window Overlap and Hop Size” on
page 120 and “Frequency grid spacing and DFT size” on page 122).
In an unsmoothed spectrogram, the horizontal center of each box is
located at the time that is halfway through the set of samples from which
that spectrum was made (see “Time alignment of spectrogram data” on
page 132). The vertical center is located at the center frequency of the
bandpass filter whose output power is represented by the darkness of the
box. The size and visibility of the boxes on the screen depends on the size
of the entire spectrogram on the screen, which in turn depends on the size
of the sound window and the signal length (which affects the time
dimension only), and on the display scale (determined by the zoom
controls). Figure 5.20 shows an unsmoothed spectrogram of the entire
time and frequency range of a 1.8-second signal and a portion of the
spectrogram after a zoom; individual cells of the spectrogram are
135
indistinguishable in the full-scale view, but are clearly visible in the
magnified portion.
Figure 5.20. Boxy spg-two scales
Figure 5.20. Unsmoothed spectrogram views at two different
magnifications. The signal is a song of a chestnut-sided warbler. Both
views are made with a 512-sample Hann window and 50% overlap. The
upper view shows the entire song; the lower view is zoomed to the
selection. The individual spectrogram cells are clearly visible in the
lower panel.
In a smoothed spectrogram, the color of each individual display pixel is
determined by bilinear interpolation between the power values calculated
at the grid points (i.e., the centers of the boxes shown in an unsmoothed
spectrogram). Each time the spectrogram is resized, the color values for
individual pixels are recalculated. Thus no matter how much you stretch a
smoothed spectrogram, you will not see sharp-edged boxes as you would
with an unsmoothed spectrogram. In a smoothed spectrogram, the “no
data” color is displayed before the time of the first spectrum, and after the
time of the last spectrum because Raven cannot interpolate power values
beyond these points.
Spectrogram smoothing is not a substitute for the finer spectrogram grid
resolution obtained by increasing the window overlap and DFT size,
however. Both may make a spectrogram more esthetically pleasing, but
only finer grid resolution will reveal some structural details of the signal
that are invisible in a low-resolution spectrogram (Figure 5.21).
136
Figure 5.21. Low-res spg boxy/smooth vs hi-res
Figure 5.21. Difference between smoothed spectrogram calculated on
low-resolution time grid and unsmoothed spectrogram calculated on a
higher-resolution time grid. The signal is part of a song from a Cassin’s
kingbird, digitized at 44.1 kHz. All three spectrograms use a 300sample Hann window. (a) Unsmoothed spectrogram on default time
grid. Time grid spacing = 3.4 mS = 150 samples (50% window overlap).
(b) Same spectrogram parameters as (a), but with smoothed display.
(c) Unsmoothed spectrogram with higher-resolution time grid. Time grid
spacing = .68 mS = 30 samples (90% window overlap). The rapid
periodic frequency modulation is more clearly represented in (c) than in
(a) or (b).
Spectrogram slice views
Time position of a A spectrogram slice view, like a spectrogram view, has a time axis and a
spectrogram slice time position. Unlike a spectrogram or waveform view, however, the time
view axis of a spectrogram slice view is not displayed. In order to change the
time position of a spectrogram slice view, its time position must be linked
to a waveform or spectrogram view (see “Linking and unlinking views” in
Chapter 3, page 62). You can then change the time position of the
spectrogram slice view by moving the time position marker of the linked
waveform or spectrogram view.
If the spectrogram slice view’s time position is before the spectrum time of
its first spectrum (see “Time alignment of spectrogram data” on page 132),
Raven displays a message in the view pane indicating that there are no
137
spectrogram data available at the current time position (Figure 5.22). Since
Raven by default sets the time position of a view to the start of the sound
when a sound file is first opened, this message is often shown when a
spectrogram slice view is first displayed. To see valid data in the slice
view, simply move the time position marker in any linked waveform or
spectrogram view.
Figure 5.22. No slice data
Figure 5.22. Three views all linked by time position. No data are
available for display in a spectrogram slice view when the time position
is at the very start of the signal, before the time of the first spectrum in
the underlying spectrogram.
Significance of the The power values shown at each frequency in a spectrogram slice view
spectrum values displayed are expressed in decibels relative to an arbitrary power value of
1. The numeric values for frequency and relative power level at the
frequency where the mouse pointer is located are displayed in hertz (Hz)
138
and decibels (dB) in the mouse measurement field at the bottom of the
signal window.
Raven 1.4 does not display calibrated sound amplitude measurements. In other words, the dB values given by Raven do
not represent absolute sound levels. Raven’s amplitude measurements can be used to obtain accurate relative measurements within signals.
Smoothed vs. As with spectrogram views, Raven can display a spectrogram slice view
unsmoothed either smoothed or unsmoothed. An unsmoothed spectrogram slice view
display displays each data point as a horizontal line whose length is determined
by frequency grid spacing. In contrast, a smoothed spectrogram slice view
interpolates between data points and calculates a power value for each
pixel accordingly. To switch beteween these two modes, use the Smooth
Spectrogram Slice checkbox in the view’s contextual menu or in the view
menu when the spectrogram slice view is active.
Selection spectrum views
Raven updates selection spectrum views whenever the active selection
changes. If there is no active selection, or if the active selection is shorter
than the window size specified in the current view parameters, selection
spectrum views display a text message indicating why no spectrum is
shown.
By default, selection spectra are shown in red, to distinguish them from
spectrogram slice views (shown in blue by default). You can specify a
different color for selection spectrum views by editing the color scheme as
discussed in “Editing color schemes” in Chapter 3 (page 71).
Significance of the The power values shown at each frequency in a selection spectrum view
spectrum values displayed are expressed in decibels relative to an arbitrary power value of
1. The numeric values for frequency and relative power level at the
frequency where the mouse pointer is located are displayed in hertz (Hz)
and decibels (dB) in the mouse measurement field at the bottom of the
signal window.
Smoothed vs As for spectrogram views, Raven can display a selection spectrum view
unsmoothed either smoothed or unsmoothed. An unsmoothed spectrum selection view
display displays each data point as a horizontal line whose length is determined
by frequency grid spacing. In contrast, a smoothed spectrogram slice view
interpolates between data points and calculates a power value for each
pixel accordingly. To switch beteween these two modes, use the Smooth
Spectrum checkbox in the view’s contextual menu or in the view menu
when the selection spectrum view is active.
139
Changing parameters for an existing spectrographic view
To change any of the parameters of a spectrogram, spectrogram slice, or
selection spectrum view, choose Configure View... from the view’s
contextual menu or from the View menu when the view is active. You can
then enter new parameters into the dialog box that appears, then click
Apply or OK.
140
Chapter 6
Selections:
Measurements, Annotations, & Editing
About this Chapter
This chapter contains information on making selections, measurements,
and editing within sound files in Raven. In particular, you’ll learn more
about:
•
•
•
•
•
•
•
•
•
selection basics such as creating, committing, deactivating, and clearing
editing sounds using filters and amplification
selection tables
working with selections
measurements
measurement formats, precision, and presets
exporting samples from various views to text files
annotations
pitch tracking
Selection Basics
About selections A selection is a specified portion of a signal with a start time and duration,
and a lower and upper frequency. You can define a selection by clicking
and dragging the mouse within a view. (Other ways of defining and
modifying selections are discussed below.) Selections are shown in views
by colored rectangles. In earlier chapters of this manual you have seen
how to select part of a waveform or spectrogram and play or zoom to that
selection. Selections are also used to identify parts of signals that you want
to measure, edit, or save.
Selections can either be range selections or point selections. A range selection
identifies a continuous range of times and/or frequencies. A point
selection identifies a single point in time, and optionally a single
frequency. You can define range or point selections in any type of view.
Although you initially define a selection in a particular view,
selections are associated with a sound, not with any particular view, and are displayed in all views.
141
Chapter 6: Selection, Measurements, & Editing
Creating and There are two ways to create a range selection. One way, as just
modifying range mentioned, is to click and drag the mouse pointer across the portion of a
selections view that you want to select. The other way to create a range selection is to
click once at one edge or corner of the selection, and then shift-click at the
opposite edge or corner. In a waveform, all of the time between the initial
click and the shift-click will be selected. In a spectrogram, a rectangular
region will be selected, with the locations of the click and shift-click
defining diagonally opposite corners. Both of these methods require that
you be in Create Selection Mode (cursor shaped like crosshairs). You
cannot create or modify selections from either of the other selection
modes.
For selections that extend beyond the limits of the current window, Raven
will auto-scroll if you drag the mouse outside the sound window.
Alternatively, Shift-clicking provides a convenient way to create long
selections. Click once at the beginning of the section, and then use the
scroll thumb to move the view so that the end of the desired section is
visible, and shift-click there. The entire section will be selected, even
though one end of the selection is beyond the edge of the window.
When you first create a selection, the selection is displayed with a dashed
border that is red by default. (You can change the color used to draw new
selections, as described in “Editing color schemes” in Chapter 3 (page 71).)
By default, a selection label appears near the top left corner of the selection,
displaying the selection ID number.
Selection control points
As soon as you release the mouse button, selection control points appear at
the edges and at the center of the selection. In spectrogram views,
selection control points also appear at the corners of a selection. Selection
control points are small squares that you can grab and move with the
mouse to adjust the boundaries of the selection or move the selection
(Figure 6.1). When you move the mouse pointer over a selection control
point, the mouse pointer changes to an icon indicating the directions that
you can move that control point. At any given time, one control point may
be active. The active control point is shown as a solid square; inactive
control points are shown as open squares. For example, the bottom right
control point in Figure 6.1 is active; the others are inactive. The active
control point can be moved using the arrow keys on the keyboard. To
activate a selection control point, click on it. Once one of a selection’s
control points is active, you can activate successive control points for that
selection by repeatedly pressing the <Tab> key.
Whenever the mouse is over a view containing an active selection, one of
the selection control points, called the extender point, will be displayed at a
slightly larger size than the others. The extender point is always the closest
control point to the current position of the mouse pointer. For example, in
Figure 6.1, the upper right corner control point is the extender point. You
can move the extender point to a new location in a view by shift-clicking at
that location.
142
Figure 6.1. Single selection, active, uncommitted
Figure 6.1. An active, uncommitted range selection. The bottom right
control point is active. The top right control point is the extender point.
Committed vs. uncommitted selections
When you first create a selection, it is uncommitted. An uncommitted
selection is temporary— it disappears as soon as you click somewhere else
in one of the sound window’s views to create a new selection. Thus there
can only be one uncommitted selection in a selection table at any one time.
To commit a selection, press the <Enter> key when any one of the
selection’s control points is active. Once a selection is committed, it will
persist in the sound window until you explicitly clear it. When you
commit a selection, its border changes from a dashed to a solid line.
The active selection
When you first create a selection, it is active. After a selection has been
committed, it becomes inactive when you define a new selection (which
then becomes active). Thus, only one selection (at most) can be the active
selection. Active and inactive selections are highlighted in different colors.
By default, the active selection is highlighted in red; inactive selections are
highlighted in cyan (Figure 6.2). Only the active selection has control
points. Certain operations— such as playback, zoom, and editing— apply
only to the active selection. To activate an inactive selection, right-click on
the selection (or Control+click on a Mac) and choose Selection N > Activate
(where N is the selection number) from its contextual menu. You can also
use Activate Selection Mode (cursor resembles a pointing hand) to activate
selections in the current selection table when you click on them.
143
Figure 6.2. Multiple selections on spg.
Figure 6.2. A spectrogram showing multiple selections. The committed
active selection is outlined in red; inactive selections are outlined in
cyan.
Point selections A point selection is created by clicking (without dragging the mouse) in a
view. A point selection is always shown in a waveform view as a colored
vertical line. The appearance of a point selection in a spectrogram or
spectrogram slice view depends on which type of view was clicked to
create the selection. If a point selection is created by clicking on a
waveform, it is shown on spectrogram views as a colored vertical line
(Figure 6.3). This is because a point selection created in a waveform
includes all frequencies by default. In a spectrogram slice view, such a
selection would appear as a highlighted rectangle covering the entire
spectrum that is visible only when the view’s time position is at the time of
the selection.
If a point selection is created by clicking on a spectrogram or spectrogram
slice view, it’s shown as a colored ‘+’ symbol in spectrogram views, and as
a vertical line in spectrogram slice and waveform views (Figure 6.3).
A point selection is visible in a spectrogram slice view only if
the view is positioned at the time of the selection.
144
Figure 6.3. Point selections.
Figure 6.3. Point selections (in green) as they appear in different view
types. Time positions of all three views are linked to each other.
Selection 1 was created by clicking in the waveform. Selection 2 was
created by clicking in the spectrogram. Selections 1 and 2 are not
visible in the spectrogram slice view because the time position is not at
the time of either selection. Selection 3 was made by clicking in the
spectrogram slice view. In the waveform, selection 3 is hidden behind
the time position marker.
Deactivating and To deactivate the active selection, choose Selection N > Deactivate from
clearing selections the selection’s contextual menu, or activate another selection. To remove
the active selection, choose Active Selection > Clear from the selection’s
contextual menu or press the Clear Active Selection button on the View
toolbar. To remove an inactive selection, choose Selection N > Clear from
the selection’s contextual menu. You can also remove all selections, or all
selections in a given table, by choosing Clear All Selections (all Tables) or
Clear All Selections (Table N) either from the View menu, from the
145
contextual menu for any selection, or by pressing the Clear All Selections
(all Tables) or Clear All Selections (Table N) button on the View toolbar.
To clear a selection means to remove the selection, leaving
the data in it untouched. To delete a selection means to
delete the data identified by that selection; in the process the
selection itself is cleared.
Renumbering When you initially create selections, Raven assigns selection ID number in
selections the order that you create the selections, beginning at 1. To renumber
selections beginning at 1, in the order determined by the current sort order
of the selection table, choose Renumber Selections from the selection
table’s contextual menu.
Renumbering is useful if the order in which selections were
created is not the order in which they occur in the sound, or if
some selections were cleared, creating gaps in the numbering sequence. Sort the table by Begin Time, then renumber to
create an uninterrupted series of selection IDs in time order.
Editing a sound
You can delete the data in the active selection either by choosing Edit >
Delete, by pressing the Delete button, by pressing <Ctrl-Delete>
(Windows, Linux) or <Command-Delete> (Mac OS), or by pressing the
Delete key on your keyboard when a selection control point is active.
Cut, Copy, Paste, and Delete operations apply only to the time dimension
of a sound window. If you select a region for one of these operations in a
spectrogram, the operation applies across all frequencies in the signal for
the duration of the selection, irrespective of the frequency limits of the
selection. Also, the clipboard can only hold one item at a time; whenever
you Cut or Copy, the data that go into the clipboard replace what was
there before.
When you modify a sound, Raven adds an asterisk (*) to the beginning of
the title of the sound window to show that you’ve edited the sound. If the
file is restricted to read-only access, the title bar tells you that, too.
If you perform Cut, Paste, or Delete operations in a sound
window that contains selections at times later than the time
where the editing operation is performed, those selections will
no longer refer to the same data as they did before the editing
operation.
146
Undoing changes The most recent change to a sound window can be undone by choosing
Edit > Undo or by typing <Ctrl-Z> (Windows, Linux) or <Command-Z>
(Mac OS). By repeatedly using the Undo command, you can undo multiple
changes, from most to least recent. You can redo a change you’ve undone
by choosing Edit > Redo or by typing <Ctrl-shift-Z> (Windows, Linux) or
<Command-shift-Z> (Mac OS). You can redo, one at a time, as many
changes as you’ve undone. By default, you can undo (and then redo) up to
five changes. You can change this number by altering the user preferences
(see “About Raven preferences” in Chapter 11 (page 299) for more
information).
Filtering and Raven enables you to selectively filter out all energy in particular
amplifying sounds frequency bands of a selection or an entire sound window. Alternatively,
you can apply an adaptive filter to reduce noise that is less concentrated in
a particular frequency range, or exists in a similar range to the desired
signal. You can also amplify all or part of a sound by any specified factor,
including factors less than 1. Amplifying by a factor less than 1 attenuates
the sound.
Band Filtering Raven allows you to filter out or filter around a selected frequency band in a
sound. When you filter out a frequency band, frequency components in
that band are removed from the signal. When you filter around a
frequency band, frequency components outside of that band are removed,
leaving only the frequencies in the selected band. The filtering operation
can be performed on the entire signal or on a time interval that you select.
To filter out or around a particular frequency band, make a spectrogram of
the sound, and then use the mouse to select the frequency band to be
filtered. To filter the selected frequency band over just the selected time
interval, choose Edit > Band Filter > Around Active Selection or Edit >
Band Filter > Out Active Selection. To filter the selected frequency band
over the entire length of the sound, choose Edit > Band Filter > Around All
or Edit > Band Filter > Out All. If there is no active selection, the Edit >
Band Filter menu is not available.
Specifically, Raven uses the Window method for FIR filter
design. A Kaiser window is used with a default transition
bandwidth of 0.02 times the Nyquist frequency and a default
stop band attenuation of 100 dB. For a complete description
of this method, see Discrete-Time Signal Processing (Second
Edition), by Alan Oppenheim, Ronald Schafer, and John
Buck, Prentice Hall 1998, pp. 474-476.
Defining your own Raven allows you to save named presets for bandpass or bandstop filters
filters covering a specified frequency band. To define a new filter, choose Edit >
Band Filter > Active Selection With > New Filter... or Edit > Band Filter > All
With > New Filter... In the ensuing dialog box, you can specify Bandpass or
Bandstop and the filter limits, and optionally save the filter settings as a
147
preset. To apply a saved filter preset to the active selection or to the entire
sound, choose Edit > Band Filter > Active Selection With >
filterPresetName or Edit > Band Filter > All With > filterPresetName. If you
choose to create your own filter, you can enter your own frequency band
information in the dialog box that appears (see Figure 6.4).
Figure 6.4 Create filter
Figure 6.4. The Create New Filter dialog box. In this case, we chose to
define a filter to work on an active selection.
Adaptive Filtering Unlike a band filter, which requires specific information regarding which
regions to filter, an adaptive filter continually modifies its parameters to
produce the best approximation of the actual signal. This is particularly
useful for signals where the background noise is not confined to a
particular frequency band, or is distributed amidst the desired signal.
Selecting Narrowband or Broadband output in the configuration dialog
specifies whether to isolate a narrowband signal within broadband noise,
or remove narrowband interference from a broadband signal.
The parameters in the upper portion of the configuration dialog affect
how the filter adapts to changes in the sound. For most purposes, the
default parameters work well. Adjusting the filter order or LMS step size
may result in a more heavily filtered output, but risks filtering out some of
the desired signal as well. When the ALE delay parameter is set to 1, the
adaptive filter functions as a classic whitening filter, separating out the
white noise component of the sound.
148
Figure 6.5Configure ALE
Figure 6.5. The Adaptive Line Enhancer dialog box, using default
parameters.
Amplifying To amplify all or part of a sound, choose Edit > Amplify... . The Amplify
Sound dialog box that appears (Figure 6.6) lets you choose whether to
amplify the entire sound or just the active selection. You can also choose
between four methods of amplification: you can either specify a factor by
which Raven will multiply each sample in the sound, a number of decibels
to add to each sample, or that Raven should multiply the entire sound or
the active selection by whatever factor necessary to make the RMS
amplitude or Peak Amplitude of the active selection equal to a specified
value.
To reduce the amplitude of a sound, amplify by a factor
greater than 0 and less than 1. Amplifying by 0 will silence the
chosen portion of a sound.
149
Figure 6.6. Amplify dialog
Figure 6.6. The Amplify Sound dialog.
Batch Filtering and The Batch Filter and Batch Amplify functions allow you to filter or amplify
Amplifying all sounds in a given folder. For example, this can be useful if you want to
filter the same frequency range on many sounds. To apply a filter go to
Tools > Batch Band Filter or Tools > Batch Adaptive Filter. In the ensuing
dialog box, you can choose the input folder containing the files you wish
to filter, and an output folder where you would like the filtered files to be
saved. This dialog also allows you to specify the type of filter and the filter
parameters. To amplify all sounds in a folder go to Tools > Batch Amplify.
In this dialog box, you can specify the input and output folders as for a
batch filter and choose how the files should be amplified (multiplying by a
given factor, adding a certain number of decibels, or setting the peak or
RMS amplitude to a given value).
Selection Tables
The Selection Table At the bottom of each sound window is a selection table, which by default is
collapsed so that it’s not visible. To expand the selection table so that it’s
visible, click on the upward-pointing Expand Selection Table button at
the left (Windows, Linux) or right (Mac OS) end of the textured separator
bar along the bottom edge of the sound window. If you click on the same
button again, the selection table will expand to occupy the entire window,
hiding the window’s views. Clicking on the downward-pointing Collapse
Selection Table button reduces the size of the selection table. You can also
drag the textured separator bar up or down to open or resize the selection
table.
By default, the selection table displays one row for each selection in each
view of the sound (Figure 6.7). Raven assigns a sequential selection ID,
150
beginning at 1, to each selection that is defined for a sound window. Each
selection’s ID is shown in the first column of the selection table, along with
a colored square that indicates whether or not the selection is the active
selection (by default, red for the active selection, cyan for all others). The
second column, labeled “View”, displays an icon to indicate whether the
row refers to a waveform, spectrogram, spectrogram slice, or selection
spectrum view. (The icons for the view types are the same ones used on
the New View buttons in the view toolbar, Figure 2.10, page 35.) The
number next to the icon indicates which view of a particular type the entry
refers to, for cases where more than one view of a given type exists.
The remaining columns in each row display measurements for the given
selection and view.
151
Figure 6.7. Spg and selection table.
Figure 6.7. A sound window containing one waveform and one
spectrogram view, with multiple selections, and the corresponding
selection table.
Each sound window may have multiple selection tables.
They will be tabbed across the bottom of the sound window,
and each table operates independently and can be edited as
such.
Using selection Unless otherwise specified, initial selections made within a sound window
tables will be part of a default selection table. New selection table(s) can be
created by selecting File > New > Selection Table from the menu. This will
create a new selection table with the default measurements. To create a
new selection table that includes the same measurements as the current
table, choose File > New > Selection Table (with measurements). Running
a detector on the data will also create a new selection table (see
152
“Displaying detector information” in Chapter 10 (page 257) for more
information).
When a new selection table is created, Raven will automatically switch to make that table active. If you would prefer to
have the previous table remain active when a new one is created, change the entry in the preferences file (raven.ui.selectionTable.setNewTableActive=true) from true to false. For
more information on making changes to the preferences file,
see “About Raven preferences” in Chapter 11 (page 299).
Also from the File menu, you can choose to close the active selection table
(File > Close Selection Table) or save the table (File > Save Selection
Table “[Table name]” or File > Save Selection Table “[Table name]” As...).
You can also close a selection table by clicking the x in the left corner of its
tab.
Selection tables appear in tabbed format, and each table has a context
menu affiliated with it. To access the context menu, position the mouse
cursor over the table’s tab and right-click (or Control+click on a Mac). You
can rename, remove, change the color of, or save the table. You can also
open a new table or merge tables.
There is a context menu for the tables’ contents as well. Right-click while
the mouse is over the table (or Control+click on a Mac) and the menu will
appear. The menu has options for you to edit columns and rows, edit cell
contents, and even edit selections that appear in the table.
Visibility of view The selection table displays a separate row for each visible view of each
and channel entries visible channel of the signal (Figure 6.8). To hide entries for a view or a
channel in the table, hide the display of that view or channel by
unchecking its box in the side panel’s Layout tab.
Selections in each table can be different colors and can be shown or hidden
independently. The Draw checkbox just above the selection table controls
whether selections in that particular selection table are drawn or not
drawn in the sound window’s view(s). You can also configure the visibility
of multiple tables at once by choosing Configure Selection Visibility from
the sound window contextual menu or View menu and selecting the tables
that you wish to draw. To draw all or none of the selection tables, use the
select all or deselect all icons in the Configure Selection Visibility dialog or
to the right of the Draw checkbox. Also, the arrows to the far right of the
Draw checkbox provide an easy way to scroll through the selections
contained in each table (see Figure 6.8).
153
Figure 6.8. Selection Table, 2 views
Figure 6.8. A signal window showing three selections for two views in
one selection table (Table 1). Note also the Draw checkbox as well as
the selection scroll arrows in the upper-right corner of the table.
Sort order of view By default, Raven sorts the entries in a selection table first by selection
and channel entries number, then by view, and finally by channel. This default sort order is
illustrated for a single-channel signal in Figure 6.8. You can choose a
different sort order (e.g., by View, Channel, Selection) from the Sort Rows
By menu in the selection table’s contextual menu. Figure 6.9 shows
selections sorted by view, then by selection for a single-channel signal.
154
Figure 6.9. Selection table by view, selection
Figure 6.9. The same selections shown in Figure 6.8, sorted by view,
then by selection.
You can also choose to have the selection table sorted in ascending order
of any one of the displayed measurements or annotations by clicking on
the column header for that measurement or annotation. To sort in
descending order, shift-click on the header. To sort by multiple
measurements, click on one column header and then a second. The
selection table will be sorted by the second measurement, and if any rows
have the same value for the second measurement, they will be sorted by
the first measurement.
Saving the Choose File > Save Selection Table “[Table Name]” As... to save the
selection table to a measurements that are visible in the selection table as a tab-delimited text
text file file, for easy import into statistical, spreadsheet, or other programs. You
155
can also retrieve selection information later within Raven, as described in
the following section.
If you try to save a selection table file that does not include
values for Begin Time, End Time, Low Frequency, and
High Frequency, Raven will warn you that you won’t be
able to retrieve the selections later, and will ask if you want to
proceed.
If the selection table contains any unsaved changes, Raven displays an
asterisk (*) at the left end of the selection table’s title bar.
Retrieving Once a selection file has been saved, you can retrieve the selections by
selections from file choosing File > Open Selection Table... or File > Open Recent Selection
Table. The selections in the chosen file will be opened in a selection table
that contains columns for any measurements saved with the file.
If one or more selections already exist in the active sound window when
you retrieve selections from a file, Raven will ask you to choose whether to
(a) discard the old selections before retrieving, (b) merge the existing
selections and the retrieved selections, (c) add a new table to store the new
selections, or (d) cancel the retrieval. If you choose to merge selections, the
retrieved selections will be assigned new selection numbers if the saved
numbers conflict with existing ones. However, if you choose to open
multiple selection files simultaneously, Raven will automatically create a
new selection table for each selection file.
When Raven retrieves selections from a file, it uses the Begin Time, End
Time, Low Frequency, and High Frequency values read from the file
to create new selections in the active signal. These four measurements are
collectively known as the selection bounds.
The file from which selection bounds are retrieved need not have been
created by saving measurements in Raven. Any tab-delimited text file can
be opened as a selection table, as long as it has the right format (see
below). This means that selection files can be created manually or by an
automated process, such as a signal detection or classification program
that is not part of Raven.
Format of selection In order for Raven to retrieve selections from a file, the first line of the file
files must be a header row that contains column headings for Begin Time,
End Time, Low Frequency, and High Frequency, separated from
each other by the <tab> character. These columns can appear in the file in
any order. If the file also contains a Selection column, Raven will assign
the selection numbers saved in that column; otherwise, the imported
selections are numbered consecutively, starting at 1. Column headers for
other measurements may be present or absent; Raven ignores values in
other columns.
156
Subsequent lines in the file must contain valid entries for each of the
columns identified in the header row, separated from each other by
<tab> characters.
If any of the selection bounds given in a selection file is outside the limits
of the active signal (e.g., a time beyond the end of the signal, or a
frequency greater than the Nyquist value), Raven will alert you to this and
ask whether to abort importing selections. If you choose to continue,
selections with bounds outside of the signal limits are ignored.
Splitting a selection To manage selection tables with large numbers of selections, Raven allows
table you to split a given table into multiple smaller tables. To split a selection
table, right click on the selection table’s tab and choose Split Table from the
contextual menu. This opens a dialog in which you can choose how to
split the selection table. You can choose how many selections are included
in each split and whether or not to draw selections in the newly created
tables. From this dialog, you can choose how to name the created tables or
files, and choose a folder in which to place saved files.
If you choose to draw the new selection tables, Raven
searches through all the tables and redraws selections each
time a sound is paged. This can slow down the paging process significantly. Therefore, when working with large numbers of selections, consider choosing not to draw selections
in the new tables automatically; you can always choose to
draw these selections later from the sound window.
Working with selections
Activating a To activate a selection, click on any one of its rows in the selection table, or
selection from the choose Selection N > Activate from the row’s contextual menu. If the
selection table newly activated selection is not visible in the active view, the view’s time
position is moved so that the selection appears in the center of the view. In
paged windows (see “Paging” in Chapter 7 (page 187)), if the selection is
not presently in memory, the necessary pages of the sound will
automatically be loaded, and the selection activated. You can activate
successive selections in the table (according to the current sort order)
using either the Activate Next Selection (down) and Activate Previous
Selection (up) arrow buttons in the table’s title bar or the up and down
arrow keys.
Auto-advancing As opposed to repeatedly clicking the Activate Next Selection button, you
through selections can choose to auto-advance through selections in a table at a proscribed
rate. Each time it advances, Raven waits the amount of time specified in
the delay field in the selection table’s title bar. To the left of the delay field
are buttons to Start and Stop auto-advancing, as well as a toggle button
that allows you to choose whether or not to play each selection as you
advance. When the auto-play button is set on, Raven will play each
157
selection as it is activated and then wait the proscribed amount of time.
Auto advance will advance through the selections in the order that they
appear in the table.
Viewing selections Choosing Window > Zoom Window opens a zoom window on the Raven
in the zoom desktop. The zoom window displays a magnified view of the active
Window selection and will update its contents as the active selection changes. For
more information on the zoom window, see “The zoom window” in
Chapter 3, page 60.
Finding selections If the active selection is not visible in the active view or in the selection
table, choosing Active Selection N > Find from the View menu or the
contextual menu for the view or the selection table will cause the active
view and the selection table to scroll to the active selection. To locate a
specific selection, enter the selection number into the box just above and to
the right of the selection table or choose Find Selection... from the
selection table’s contextual menu and enter the number of the selection
you wish to find. Alternatively, double clicking a row in the selection table
will activate the corresponding selection and cause the active view to
scroll to that selection.
Figure 6.10Selection tools
Figure 6.10. A selection table with multiple selections. The controls for
navigating through selections are identified.
Modifying selection There are three ways to modify the bounds (i.e., Begin Time, End Time,
bounds Low Frequency, and High Frequency) of existing selections.
158
Moving selection control points
When a selection is active, you can modify its selection bounds by
manipulating its selection control points in any view. Selection control
points can be moved either by dragging with the mouse, or using the
arrow keys on the keyboard. In order to move a control point with the
arrow keys, you must first activate the control point by clicking on it. You
can activate different control points in succession by repeatedly pressing
the <Tab> key when any control point is active.
Editing in the selection table
You can change the Begin Time, End Time, Delta Time, Low Frequency,
High Frequency, and Delta Frequency values for a selection by editing the
value directly in the selection table. Any change that you make to one of
these fields takes effect when you press <Enter>, <Tab>, or any of the
arrow keys. If you change Delta Time or Delta Frequency field, Raven
keeps the Start Time and Low Frequency fields constant, and adjusts the
End Time or High Frequency as necessary.
Using the Selection tab in the side panel
The Selection tab in the side panel displays the Begin Time, End Time,
Delta Time, Low Frequency, High Frequency, and Delta Frequency for the
active selection (Figure 6.11). All six of these fields are editable, so you can
adjust the selection bounds by typing values into these fields. Any change
that you make to one of these fields takes effect when you press <Enter> or
<Tab>, or when you click anywhere outside of the field that you’ve
changed. If you change Delta Time or Delta Frequency field, Raven keeps
the Start Time and Low Frequency fields constant, and adjusts the End
Time or High Frequency as necessary.
159
Figure 6.11. Selection side panel
Figure 6.11. The Selection tab in the side panel. This example is from a
four-channel sound file.
Modifying selection For multi-channel signals, the Selection tab also displays a list of the
channels channels in the signal (Figure 6.11), with checkboxes indicating the
channels in which the active selection is defined. To change the channels
in which the active selection is defined, check or uncheck channels as
needed.
Selection tables in Paged sound windows provide a mechanism for working with sounds
paged sound that are too large to fit in the memory available to Raven. In a paged
windows window, only part of the entire sound, called a page, is in memory at any
given moment. Paged sound windows are discussed in detail in “Paging”
in Chapter 7 (page 187). The selection table in a paged sound window
displays all of the selections in the signal irrespective of whether the
selections are presently in memory, with their associated measurements
and annotations. If you activate a selection in the table for which the audio
data are not in memory, the page centered around the selection is loaded
(replacing the current page), and the active view is centered around the
selection.
Saving the active To save the audio data in the active selection to a separate file, choose File
selection > Save Active Selection As... or use the Save Active Selection button in
the file toolbar.
160
Saving all To save the audio data in all selections to a series of separate sound files,
selections choose File > Save All Selections In Current Table As... or use the Save All
Selections button in the file toolbar. Raven will display the Save All
Selections dialog box, which allows you to specify a file format and a
naming scheme for the files to be saved.
Note that when you save the audio data in all selections, Raven will also
create a list file, which is a text file listing all of the selections’ file names.
This list file can be used to open all of the selections at once in a file
sequence. You can specify the list file’s name in the File Names tab of the
Save All Selections dialog. The File Names and File Format tabs in this
dialog work in the same way as those used in the New Recorder dialog,
with a few small exceptions. For more information, see “Recording to a file
sequence” in Chapter 4 (page 92). When saving selections, you can include
information from annotation columns in the file name by including <a> in
the name and selecting the appropriate column name(s) from the list
below. Also, you can set a non-zero pad length to include data up to the
specified time before and after each selection in the exported clips.
If a background task such as a detector is running, saving
selections or selection tables is temporarily disabled. This
prevents Raven from attempting to save selections from a
selection table that is being updated by a detector.
Saving all selections operates as a background task in Raven, allowing you
to continue working in Raven during this time. However, some functions
are disabled during this process to prevent Raven from modifying a
selection that is being saved. You cannot add, delete, or modify entries in
the relevant selection table, or run a detector while saving selections. You
can, however, continue to work with other selection tables or sounds
during this time.
Highlighting To highlight multiple rows in a selection table, highlight the first row in
selections the range, and then select the last row in the range while holding down
the <shift> key. Alternatively you can choose Highlight Rows... from the
View menu or right click on the selection table and choose Highlight
Rows... from the contextual menu. You can then enter the IDs of the first
and last selections in the range.
Copying selected You can copy selected rows from a window’s selection table, and paste the
rows from the measurements into another program, such as a spreadsheet, word
selection table processing, or statistics program.
To select a single row for copying, click anywhere in the row. To select a
range of rows, click on the first (or last) row you want, then shift-click on
the last (or first) row you want. To select multiple rows that are not
adjacent to each other, hold down the <Ctrl> (Windows, Linux) or
161
<Command> (Mac OS) key while clicking on additional rows after
selecting the first row. Rows that are selected are highlighted in color.
To copy the selected rows, choose Copy Selected Rows from the selection
table’s contextual menu, or press <Ctrl-Shift-C> (Windows, Linux) or
<Command-Shift-C> (Mac OS).
Copying or moving Selections can be moved between selection tables within a sound. To move
selections between selections, select their corresponding rows in the selection table, right-click
selection tables on those rows (or Control+click on a Mac), and choose Move Selection(s)
To > [Table name] from the selection’s context menu to move the
selections to the chosen table. You can also choose Move Selection(s) To >
New Table to create a new table with these selections. To copy selections to
a different table, choose Copy Selections To > [Table name] or Copy
Selections To > New Table from the selection’s context menu.
When you move selections between tables, only the measurements
configured in the destination table will be displayed. To display additional
measurements, add these to the new table by choosing View > Choose
Measurements or right clicking on the selection table tab and selecting
Choose Measurements from the table’s contextual menu. You can also
configure Raven to add an annotation column with the source table name
in it. To do this, uncomment the preferences entry
(Raven.ui.selectionTable.storeSourceTableNameOnCopyOrMoveSelection
.annotationColumnName=) and add a name for the new column to the
right of the equals sign.
For more information about editing the Raven preferences file, see “About
Raven preferences” in Chapter 11 (page 299).
Clearing selections To clear a selection means to remove the selection information, without
modifying the audio data contained in the selection. In contrast, to delete a
selection means to delete the audio data contained in the selection, as well
as the selection information.
To clear the active selection, do one of the following:
• Choose View > Active Selection N > Clear.
• In any view, choose Active Selection N > Clear from the selection’s
context menu.
• In the selection table, choose Clear Selection N from the context menu
for any of the selection’s rows.
• Press the Clear Active Selection button in the file toolbar.
To clear a single inactive selection, do one of the following:
• In any view, choose Selection N > Clear from the selection’s context
menu.
• In the selection table, choose Clear Selection N from the context menu
for any of the selection’s rows.
162
To clear multiple selections at once, highlight rows from those selections
in the selection table, then choose Clear Selections from the context menu
for any of the highlighted rows. To highlight a series of contiguous rows,
click on the row at one end of the series (i.e., the first or last row of the
series), then shift-click on the row at the other end of the series. To
highlight multiple non-contiguous rows, click on the first row to be
highlighted, then control-click on other rows.
To clear all selections in a signal choose Clear All Selections from the
View menu, from the contextual menu for any view or selection, from any
row in the selection table, or by pressing the Clear All Selections button in
the file toolbar.
Splitting a selection To split a selection into multiple shorter back-to-back selections, activate
the selection you’d like to split, then choose View > Active Selection N >
Split. This will bring up the Split Selection dialog (Figure 6.12).
Figure 6.12
Figure 6.12. The Split Selection dialog.
Choose one of the six selection splitting options in the dialog. The three
options in the Time Domain section will split the selection into vertical
bands, and the three options in Frequency Domain will split the selection
into horizontal bands (Figure 6.13).
163
• Equal Duration: split the selection evenly into the given number of
smaller vertical selections.
• Equal Power: split the selection into the given number of smaller vertical selections, each of which may be a different width but each of which
will contain an equal total power.
• Fixed Duration: split the selection into smaller vertical selections of the
given duration. If the duration does not fit evenly into the length of the
entire selection, the last selection will be truncated.
• Equal-size Bands: split the selection evenly into the given number of
smaller horizontal selections.
• Equal Power: split the selection into the given number of smaller horizontal selections, each of which may be a different height but each of
which will contain an equal total power.
• Fixed-size Bands: split the selection into smaller horizontal selections
of the given bandwidth. If the bandwidth does not fit evenly into the
height of the entire selection, the last selection will be truncated.
If you check Remove Original Selection, the original large selection will be
replaced by the smaller selections. If you don’t check it, the original large
selection will remain, superimposed by the smaller selections.
164
Figure 6.13
Figure 6.13. Examples of the six options for splitting a selection. a)
The equal duration option in the Time Domain, set to generate 10
selections. b) The equal power option in the Time Domain, set to generate 10 selections. c) The fixed duration option in the Time Domain,
set to generate 0.05-second selections. d) The equal-size bands
option in the Frequency Domain, set to generate 10 selections. e) The
equal power option in the Frequency Domain, set to generate 10 selections. f) The fixed-size bands option in the Frequency Domain, set to
generate 1000-Hz selections.
Generating multiple You can generate multiple back-to-back selections by creating one large
back-to-back selection over the relevant area, then splitting the selection (see “Splitting
selections a selection” in Chapter 6, page 163). Alternately, you can choose View >
Generate Selections. This will bring up the Generate Selections dialog
(Figure 6.14).
165
Figure 6.14
Figure 6.14. The Generate Selections dialog.
Enter a start and end time for the area where selections will be generated,
then choose Equal Duration or Fixed Duration. For more information
about those options, see “Splitting a selection” in Chapter 6 (page 163).
Creating duplicate To create a selection identical to the active selection, but located at a
selections at different time, first choose View > Active Selection N > Copy Bounds.
different times Then position the mouse pointer at the point in time where you want the
new selection to be centered, right-click (Windows, Linux) or Control-click
(Mac OS) to and choose Paste Selection Bounds from the contextual
menu.
Cloning selections To create a duplicate selection that is superimposed upon the active
selection choose View > Active Selection N > Clone Bounds. The new
selection will be activated and displayed on top of the previously active
selection. You can then manipulate the new selection via its control points.
Measurements
Choosing By default, Raven displays four measurements in the selection table: Begin
measurements to Time, End Time, Low Frequency, and High Frequency. These values,
display collectively called the selection bounds, define the edges of a selection in a
spectrogram view. However, you can choose to have Raven display any
combination of measurements from a large repertoire. Choosing Choose
Measurements from the View menu or from the selection table’s
contextual menu displays the Measurement Chooser dialog box (Figure
6.15). The left list shows the measurements that are currently displayed in
the selection table. The right list shows all of the measurements that are
available in Raven. To add a measurement, select the name of the
measurement in the right list and then click the left-pointing arrow
between the two lists to add it to the left list. To remove a measurement,
166
select its name in the left list and then click the right-pointing arrow. Shiftclick to select multiple adjacent measurements in either list; control-click
to select multiple non-adjacent measurements. When you’ve finished
configuring the set of measurements you want, click OK.
The order in which measurement columns are displayed in
the selection table is determined by the order in which they
appear in the list of Displayed Measurement in the Measurement Chooser.
Figure 6.15. Choose Msmts dialog.
Figure 6.15. The Measurement Chooser dialog box.
Plottable Some of the measurements in Raven can also be plotted in the sound
Measurements views. To display these measurements, select the Enable Measurement
Plots checkbox in the Choose Measurements dialog. With this box
checked, you can then choose which measurements to plot by selecting or
deselecting the Plot checkbox. Note that not all measurements may be
plotted. A list of the plottable measurements is shown below. Descriptions
of these measurements can be found in the following sections.
•1st Quartile Frequency
•Max Power
•1st Quartile Time
•Max Time
•3rd Quartile Frequency
•Min Amplitude
•3rd Quartile Time
•Min Time
•Average Power
•Peak Amplitude
167
•Center Frequency
•Peak Correlation
•Center Time
•Peak Frequency
•Energy
•Peak Lag
•filtered RMS Amplitude
•Peak Power
•Frequency 5%
•Peak Time
•Frequency 95%
•RMS Amplitude
•Max Amplitude
•Time 5%
•Max Bearing
•Time 95%
•Max Frequency
Measurements In Raven, Power Spectral Density is computed by:
based on
spectrogram values 1. defining the signal in the time domain by selecting it
2. taking a Fourier transform to define the signal in the frequency domain
3. and squaring the magnitude of each resulting Fourier coefficient.
The resulting values are used to create the columns of gray-scale or color
values of the pixels that make up Raven’s spectrogram view. The
following measurements all make use of power spectral density, or
spectrogram, values.
Average Power
(spectrogram, spectrogram slice, selection spectrum)
The value of the spectrogram’s power spectral density, as it appears in
each pixel, or bin, of the spectrogram, averaged over the selection (that is,
the frequency-time rectangle that forms a selection in Raven). The values
of the spectrogram’s power spectral density are summed, and the result is
then divided by the number of time-frequency bins in the selection.
In the spectrogram slice view, Average Power is the value of the power
spectrum (the power spectral density of a single column of spectrogram
values) averaged over the frequency extent of the selection. The values of
the power spectrum are summed between the lower and upper frequency
bounds of the selection, and the result is divided by the number of
frequency bins in the selection. Units: dB.
Delta Power
(selection spectrum)
The difference between the value of the power spectrum (power spectral
density, in power per unit frequency) at the upper and lower frequency
limits of the selection; computed as the value of the power spectrum at the
upper frequency limit minus the value of the power spectrum at the lower
frequency limit. Units: dB.
168
Energy
The total energy within the selection bounds.
For a spectrogram, the energy is calculated as
 t2 f2


(
S
⁄
10
)
 W ⋅ 10 t, f
  Δf
Formula 6.1 
 0


 t = t1 f = f1


where f1 and f2 are the lower and upper frequency limits of the selection,
t1 and t2 are the beginning and ending frame numbers of the selection, W0
is the power dB reference value, St,f is the spectrogram power spectral
density in frame t at frequency f (in dB), and Δf is the frequency bin size
(which is equal to the sample rate divided by the DFT size). In Raven, the
reference power W0 is always equal to unity.
For a spectrum, energy is calculated as
 f2


 W ⋅ 10 ( S f ⁄ 10 )  Δf
Formula 6.2 
 0


 f = f1


where f1 and f2 are the lower and upper frequency limits of the selection,
and Sf is the value of the power spectrum (power spectral density) at
frequency f, in dB. Units: dB.
Aggregate Entropy
The aggregate entropy measures the disorder in a sound by analyzing the
energy distribution within a selection. Higher entropy values correspond
to greater disorder in the sound whereas a pure tone with energy in only
one frequency bin would have zero entropy. By treating the fraction of
energy in a selection present in a given frequency bin as a probability,
Raven calculates the entropy using:
169
f2
Formula 6.3
H sel =

 E bin
 E bin 
-------------------⋅
log

2 E  
E
 sel
 sel  
f = f1
where Hsel is the aggregate entropy in the selection and f1 and f2 are the
upper and lower frequencies bounds of the selection. Ebin corresponds to
the energy in a specific frequency bin over the full time span of the
selection, and Esel is total energy summed over all frequency bins in the
selection. The size of a frequency bin is determined by the spectrogram
parameters as described in“Frequency grid spacing and DFT size” on
page 339.
Average Entropy
The average entropy in a selection is calculated by finding the entropy for
each frame in the selection and then taking the average of these values.
Unlike the aggregate entropy which uses the total energy in a frequency
bin over the full time span, the average entropy calculates an entropy
value for each slice in time and then averages. As a result, the average
entropy measurement describes the amount of disorder for a typical
spectrum within the selection, whereas the aggregate entropy corresponds
to the overall disorder in the sound.
Max Frequency/ Peak Frequency
(spectrogram, spectrogram slice)
The frequency at which Max Power/ Peak Power occurs within the
selection. If Max Power/ Peak Power occurs at more than one time and/or
frequency, the lowest frequency at Max Time at which Max Power/ Peak
Power occurs. Units: Hz.
Max Power/ Peak Power
(spectrogram, spectrogram slice)
The maximum power in the selection. In a grayscale spectrogram, the
maximum power in a selection is the power at the darkest point in the
selection. Units: dB re 1 dimensionless sample unit.
Robust signal The following eight measurements are considered robust signal
measurements measurements. They should not vary much based on the exact bounds of
the selection that a user or detector creates. Unlike the selection-based
measurements, which rely entirely on the time and frequency endpoints of
the selection, these measurements take into account the energy that is
stored in the selection; so small changes in the borders of the selection
rectangle should have little effect on the computed measurement values.
For an in-depth description of the robustness of these measurements, and
170
more detail regarding the research behind them, see http://
www.birds.cornell.edu/brp/research/algorithm/automated-and-robustmeasurement-of-signal-features/.
Center Frequency
The frequency that divides the selection into two frequency intervals of
equal energy. For the spectrogram view, the Center Frequency fc is the
smallest discrete frequency within the selection which satisfies the
inequality
fc
Formula 6.4
t2
 S
f = f1 t = t1
f2
t, f ≥
t2
 S
t, f
f = fc + 1 t = t1
where St,f is the value of the spectrogram power spectral density at
discrete time t and discrete frequency f. It can also be stated that the
Center Frequency is the smallest discrete frequency in which the left side
of the formula exceeds 50% of the total energy in the selection.
For the spectrogram slice view, the procedure is the same except that the
summation over time is not necessary since selections in the slice view
occupy only one time bin. Units: Hz.
1st Quartile Frequency
The frequency that divides the selection into two frequency intervals
containing 25% and 75% of the energy in the selection. The computation of
this measurement is similar to that of Center Frequency, except that the
summed energy has to exceed 25% of the total energy instead of 50%.
3rd Quartile Frequency
171
IQR (Inter-quartile Range) Bandwidth
The difference between the 1st and 3rd Quartile Frequencies.
Units: Hz.
Frequency 5%
(spectrogram, spectrogram slice, selection, selection spectrum)
Frequency 95%
Bandwidth 90%
The difference between the 5% and 95% frequencies.
Units: Hz
Center Time
The point in time at which the selection is divided into two time intervals
of equal energy. For the spectrogram view, the Center Time tc is the
smallest discrete time within the selection which satisfies the inequality
f2
Formula 6.5
tc
 S
f = f1 t = t1
172
f2
t, f ≥
t2

f = f1 t = tc + 1
S t, f
where St,f is the value of the spectrogram power spectral density at
discrete time t and discrete frequency f. It can also be stated that the
Center Time is the smallest discrete time in which the left side of the
formula exceeds 50% of the total energy in the selection. Note that the
accuracy of this measurement is limited to the resolution of the
spectrogram, as determined by the DFT size.
occupy only one time bin. Units: seconds.
1st Quartile Time
The point in time that divides the selection into two time intervals
this measurement is similar to that of Center Time, except that the
3rd Quartile Time
IQR (Inter-quartile Range) Duration
The difference between the 1st and 3rd Quartile Times.
Units: seconds.
Time 5%
173
Time 95%
(spectrogram, spectrogram slice, selection spectrogram)
Duration 90%
The difference between the 5% and 95% times.
Units: seconds
Measurements The following measurements are based on the amplitude values of the
based on waveform individual samples that make up the audio data.
values
Max Amplitude
(waveform)
The maximum of all the sample values in the selection. Note that this
measurement can be thrown off slightly by interpolation. The value
plotted on the edge of the selection in monotonic data is not exactly the
same as the measured Max Amplitude. Units: dimensionless sample
values.
Min Amplitude
(waveform)
The minimum of all the sample values in the selection. Units:
dimensionless sample values.
Peak Amplitude
(waveform)
The greater of the absolute values of Max Amplitude and Min Amplitude.
Units: dimensionless.
RMS Amplitude
(waveform)
The root-mean-square amplitude (sometimes called “effective
amplitude”) of the selected part of the signal. RMS amplitude is equal to
174
n
Formula 6.6

i=1
2
xi
----n
where n is the number of samples in the selection, and xi is the amplitude
(in dimensionless sample units) of the ith sample in the selection. Units:
dimensionless sample units.
Filtered RMS Amplitude
(waveform)
The root-mean-square amplitude for the selected time and frequency
range. This is equivalent to applying a bandpass filter to the selection and
then calculating the RMS amplitude for the resulting signal. Units:
dimensionless sample units.
Max Time
(all view types)
For a waveform view, the first time in the selection at which a sample with
amplitude equal to Max Amplitude occurs. For a spectrogram view, the
first time in the selection at which a spectrogram point with power equal
to Max Power/ Peak Power occurs. Units: seconds.
Min Time
(waveform)
The first time in the selection at which a sample with amplitude equal to
Min Amplitude occurs.
Peak Time
(waveform)
The first time in the selection at which a sample with amplitude equal to
Peak Amplitude occurs.
Selection-based Begin Time
measurements
(all view types)
The time at which the selection begins. Units: seconds.
Delta Frequency
(all view types)
The difference between the upper and lower frequency limits of the
selection. Units: Hz.
Delta Time
(all view types)
The difference between Begin Time and End Time for the selection. Units:
seconds.
175
End Time
(all view types)
The time at which the selection ends. Units: seconds.
High Frequency
(all view types)
The upper frequency bound of the selection. Units: Hz.
Low Frequency
(all view types)
The lower frequency bound of the selection. Units: Hz.
Other Begin File
measurements
(all view types)
Displays the name of the sound file in which a selection starts.
Begin Path
(all view types)
Displays the full path for the sound file in which a selection starts. For
instance, the path name for the BlueWhale example file would appear as
“C:/Program Files/Raven Pro 1.4/Examples/BlueWhale.aiff” in Windows.
End File
(all view types)
Displays the name of the file in which a selection ends.
End Path
(all view types)
Displays the full path for the file in which a selection ends. For instance,
the path name for the BlueWhale example file would appear as “C:/
Program Files/Raven Pro 1.4/Examples/BlueWhale.aiff” in Windows.
File Offset
(all view types)
For a file sequence, the amount of time between the beginning of the file
and the start of the selection. Units: seconds.
Length
(all view types)
The number of frames contained in a selection. For waveform views, the
number of frames equals the number of samples in a single channel. For
spectrogram and spectrogram slice views, the number of frames equals
the number of individual spectra in the selection in one channel. For
selection spectrum views, the number of frames always equals 1.
176
Max Bearing
(beamogram)
The bearing angle with the maximum power for a specific selection. The
angle will be represented by the area of highest energy in the beamogram.
Units: degrees.
Peak Correlation
(correlation)
The maximum value of a correlation plot. This measurement is analogous
to Peak Amplitude in the waveform view. Units: dimensionless.
Peak Lag
(correlation)
The time in a correlation plot at which the Peak Correlation occurs. This is
analogous to Peak Time in the waveform view except that it represents
the relative offset into the first sound file of the correlation at which the
second sound file achieves the best correlation with the first file. Units:
seconds.
Measurement The number of digits of precision displayed by default for measurement
precision and values is determined by measurement precision preferences, which you
format change by editing the Raven preference file. For general information on
editing the preference file, see “About Raven preferences” in Chapter 11
(page 299); for specific information on changing preferences for
measurement precision, see “Measurement presets” in Chapter 11 (page
302).
You can change the measurement precision and format for a given
measurement at any time by choosing Format Column... from the
contextual menu for the measurement’s column in the selection table. The
dialog box that appears lets you choose either a decimal or scientific
format for measurement values, and lets you specify the number of digits
of precision used.
Using The section “Choosing measurements to display” on page 166 describes
measurement how to use the Measurement Chooser to choose which measurements are
presets displayed in the selection table. You can use measurement presets to save
sets of measurement choices that can later be applied to a selection table
without having to add or remove one measurement at a time.
Measurement presets also save the format and precision of the
measurements that are displayed.
To save a measurement preset, first choose the set of measurements to be
displayed in the selection table, as described in “Choosing measurements
to display” on page 166. Then choose Save As... from the Preset menu in
the Measurement Chooser dialog (Figure 6.16). When the Save
177
Measurement List dialog appears, enter a name for the preset, and click
OK.
Measurement presets must be saved in the folder Presets/
Measurement List/ within the Raven program folder. You
can also create additional folders within the Measurement
List folder by clicking on the New Folder icon within the
Save Measurement List dialog. These folders will appear as
submenus in the Preset menu, with each submenu listing the
presets in the corresponding folder.
To apply a saved measurement preset to the selection table, choose the
name of the preset from the Preset menu in the Measurement Chooser
dialog, and click OK.
Figure 6.16. Measurement Chooser
Figure 6.16. The Measurement Chooser dialog box.
If you make changes to the list, order, format, or precision of
measurements displayed in a selection table to which a saved preset has
already been applied, you can save the changes under the same name by
opening the Measurement Chooser and choosing Preset > Save “PresetName”.
Exporting samples from various views to text files
A list of the numerical sample values contained in the active view of a
sound can be exported to a text file. This facilitates examination and
analysis of the data using other software applications. For instance, it
178
might be useful to export spectrographic information from Raven for
display as a 3D plot.
For waveform, spectrogram, correlation, and beamogram views, samples
are exported from the current page. To export from a segment of sound
other than what is represented on the page, update the page size. Samples
in spectrum slice views and selection spectrum views are exported from
the represented slice or active selection respectively.
In a Sound Window, click on the narrow, vertical view selection button to
the left of the view for which you want to export sample values. The color
of this vertical button will change from white to blue when highlighted.
Choose File > Export “Sound <x>” <selected view> Samples... A
suggested name, <FileName>.samples-ch<cc>.txt, appears in a dialog box.
Files can be saved in either Tab-delimited .txt or Comma-separated Value
.csv format. To finish, click Save.
Export output and The specific axis values contained in Raven views are not included in the
text file content exported text files. That is, only sample power values are exported; the
specific values on the x- and y-axes are not included. For 2-dimensional
data, the sample values are organized in a single column, with reference to
the x-axis values of the Raven view (but not the actual values) represented
vertically. Multi-dimensional data is exported as power values, but in this
case, columns represent the Raven view’s y-axis values, and rows
represent the Raven view’s x-axis values.
Annotations
In addition to the measurements provided by Raven, you can define
additional columns in the selection table to contain annotations that you
enter interactively. Each annotation column has a name that you define
(e.g., “Song Type” or “Individual ID”) and can contain any information
you want (Figure 6.17).
179
Figure 6.17. Annotated view and table
Figure 6.17. A series of selections with annotations. A single
annotation column named “Unit Type” is defined. Selection labels (see
“Selection Labels” on page 182) have been configured to display the
Selection ID and Unit Type values, separated by “/”.
Creating, renaming, To create and name a new annotation column, choose Add Annotation
and deleting Column... from the selection table’s contextual menu. Enter a name for the
annotation columns new annotation column in the dialog box that appears, and click OK.
To rename an existing annotation column, choose Rename Column from
the contextual menu for that column, enter a new name for the column,
and click OK.
To delete an annotation column from the selection table, choose Delete
Column from the column’s contextual menu.
If you delete an annotation column that contains annotations,
the annotation values will be deleted and cannot be retrieved.
Note that this is different from deleting a column containing
one of Raven’s built-in measurements, which will be recomputed automatically if you later add the column again.
Entering There are three ways to enter annotation values for selections.
annotations
180
Entering annotations when committing selections
When you commit a new selection (see “Committed vs. uncommitted
selections” on page 143), Raven by default displays the Annotate Selection
dialog box (Figure 6.18), which lets you enter values for all annotation
columns that are currently defined. If you do not want Raven to display
this dialog each time you commit a selection, uncheck the box labelled
Show this dialog whenever a selection is committed.
If you check the box labelled Use specified values as defaults before you
click OK, then Raven will automatically enter those values in the
corresponding dialog fields when annotating subsequent selections. When
Raven enters default values in this dialog box, you can either change the
values, or just press <Enter> or click OK to accept them.
Figure 6.18. Annotate selection dialog
Figure 6.18. The Annotate Selection dialog box. Two annotation
columns have been defined, named “Song Type” and “Individual ID”.
Entering annotations directly in the selection table
To enter an annotation value for a particular selection, you can click on the
annotation cell in the selection table and then type the annotation value,
followed by <Enter>.
To enter the same annotation value into a series of consecutive entries in a
selection table,
1. Enter the annotation for the first entry in the series.
2. Highlight the annotation cells for a series of consecutive rows
beginning with the row for which the annotation has been entered.
3. Choose Fill Selected Cells > Down from the contextual menu for the
annotation column, or press Ctrl-D (Windows, Linux) or Command-D
(Mac OS).
To fill a series of consecutive annotation cells with the value in the bottom
cell, choose Fill Selected Cells > Up, or press Ctrl-U (Windows, Linux) or
Command-U (Mac OS).
181
Using the Annotate Selection command
You can enter annotation values into the Annotate Selection dialog box
(Figure 6.18) for a selection by choosing Annotate from the contextual
menu that appears for the selection displayed in any view. (Depending on
whether the selection is active, the menu command will either be
displayed as Active Selection N > Annotate or Selection N > Annotate.)
Selection Labels By default, Raven displays each selection’s ID number in selection labels in
views. To toggle display of selection labels on and off, use the checkbox in
the component list of the side panel’s Layout tab (Figure 2.13, page 40).
Selection labels can display selection IDs and/or any combination of
measurements or annotations that are shown in the selection table. To
choose which measurements are included in selection labels, as well as
specifying font size and a string to use as a separator between items in
selection labels, choose Configure Selection Labels from the View menu or
any view’s contextual menu. To add a measurement or annotation to
selection labels, highlight that item’s name in the list of Available Items,
and click the left-pointing arrow button in the Configure Selection Labels
dialog box (Figure 6.19). To remove an item from selection labels,
highlight its name in the list of Displayed Items, and click the rightpointing arrow button. You can also enter a string of text that will appear
as a separator between items in the selection labels; the default separator is
a single space character.
Figure 6.19. Configure selection labels dialog
Figure 6.19. The Configure Selection Labels dialog box.
182
You can specify colors for active and inactive selection labels via the color
scheme editor (View > Color Scheme > Edit...), as described in “Editing
color schemes” in Chapter 3 (page 71).
Colors chosen in the above manner will pertain to selections
made in the first selection table. Colors for additional
selection tables can be set in the corresponding tab’s context
menu.
Pitch tracking
Raven Pro’s selections and measurements allow you to perform pitch
tracking: tracing the dominant frequency of a sound over time (see Figure
6.20). This can be useful for classifying and comparing sounds. This
section contains a walk through of how to use pitch tracking.
Figure 6.20
Figure 6.20. A spectrogram of a Bowhead Whale’s song. Two calls
have been pitch tracked: Raven has measured and highlighted their
dominant frequency over time.
First, open a sound file. Note that a spectrogram view must be visible in
order to perform pitch tracking.
Choose the segment of the sound that you will work with. This is usually
a single call of interest. If you choose a segment of sound with multiple
183
calls and periods of silence between calls, the pitch tracking results for the
periods of silence will be meaningless. In the Bowhead Whale recording
shown in Figure 6.20, we selected the period of time from 1:01.15 to
1:03.65. We later repeated the process for the period of time from 1:05.45
to 1:08.05, to pitch track the second call.
Right-click on the spectrogram and choose Generate Selections. Enter the
Start Time and End Time that you’ve chosen. Under Method, select Fixed
Duration and enter your chosen selection duration. This duration will act
as a “step size,” the length of a single peak frequency measurement.
Different durations are appropriate for different sounds; a short and
complicated call will need a smaller selection duration than a long and
simple call. See Figure 6.22 for an example of the effects of different
selection durations. In our example case, the call is several seconds long
but fairly varied in frequency, so we’ve chosen a selection duration of 0.05
seconds. Click OK and the selections will be generated (Figure 6.21)
Figure 6.21
Figure 6.21. The first Bowhead Whale call after generating selections.
The period of time we’ve selected is filled with 0.05-second selections.
From the selection table’s context menu, click Choose Measurements and
add the Peak Frequency measurement. For more information about
adding measurements, see “Choosing measurements to display” in
If you want to use the pitch tracking information as numerical data, you
can now save the sound’s selection table and use the values in the Peak
Frequency column to perform your analysis.
To create an image containing the pitch tracking information, go to the
Components section of the Layout tab on the side panel (see “Hiding and
showing window components” in Chapter 3, page 67). Check the box
184
labeled Measurements to turn on measurement plotting, and uncheck the
checkboxes labeled Selection Borders, Selection Fill, and Selection Labels
to hide the selections.
Now your spectrogram should be overlayed with lines representing the
Peak Frequency measurements. By default, the lines are dark green. If
that is not easily visible against the spectrogram color map you’ve chosen,
you can change it in the Color Scheme Manager (see “Editing color
schemes” in Chapter 11, page 289). Here, we’ve changed it to yellow for
better visibility.
Figure 6.22
Figure 6.22. The results of two pitch tracking runs. On the first call, we
used a selection duration of 0.05 seconds; on the second call, we used
a selection duration of 0.5 seconds. Note that 0.5 seconds is too long
for this call; the frequency changes are not accurately reproduced by
the pitch tracking.
185
186
Chapter 7
Large Datasets
About this Chapter
Sounds can be collected and managed in a variety of ways. Some
continuous recordings are made into sequences of large files, while others
are made into sequences of many small files. Other recordings are
separated into smaller clips of relevant information by using detectors or
browsing and making selections by hand and saving the selections into
individual files. All of these cases fall into a broad category of having a lot
of data to manage, or having a large dataset.
Raven offers many features for working with large datasets including
paging, performing batch operations, and handling streams. This chapter
includes sections on the following topics:
•Paging
•Batch operations
•Opening file sequences
•Opening multiple sound files in multiple windows
•Streams
•Opening sounds with altered speeds
Paging
Paged sound windows enable you to open sounds that are too large to fit in
the memory available to Raven. In a paged sound window, Raven loads
only a specified amount of the sound, called a page, into memory at one
time. The portion of the entire signal that is presently in memory is called
the current page. A paged sound window includes a page navigation panel
that enables you to jump easily to any other part of the signal that’s not
presently in memory. Paged sound windows can be used to display single
sound files or file sequences.
Paged sound windows behave like non-paged windows,
except that you cannot perform editing operations that alter
the audio data (i.e. Cut, Paste, Delete, Filter, or Amplify).
187
Chapter 7: Large Datasets
Configuring a new When you first open a sound file or a sound file sequence, Raven displays
paged sound the Configure New Sound Window dialog box, which gives you the
window choice of loading the entire sound into memory or creating a paged sound
window (Figure 7.1). To open a sound in a new paged sound window,
choose Page Sound. To specify how much of the sound should be loaded
into memory at any given time, enter a value into the Page size field. You
can choose units of seconds or minutes for page size from the drop-down
menu.
When you plan to browse large sections of data, you may
wish to use a page size of twice the length you plan to view at
a time. Then, before you begin browsing, zoom to half of the
page length. This helps to reduce interruptions from progress
bars as you scroll through the data.
Page increment and Step increment specify the amount by which the
paged display moves when you click the Page Forward, Page Back, Step
Forward, or Step Back controls as described in “Moving the page
incrementally: the Page and Step buttons” on page 190. You can specify
Page increment and Step increment either as a percent of the page size or
as an absolute number of seconds or minutes by choosing the appropriate
units from the drop-down menu next to each field.
Figure 7.1. Configure Sound Window dialog
Figure 7.1. The Configure New Sound Window dialog box.
188
The Configure New Sound Window dialog box also allows you to choose
a window preset to be applied to the new window when it is first created,
as discussed in “Window preset and paging” in Chapter 1 (page 4).
Navigating through When you open a single file or a file sequence in a paged sound window, a
signals in paged page navigation panel is displayed at the top of the window, above any
sound windows views that are displayed (Figure 7.2).
Figure 7.2. Paged sound window-- file sequence
Figure 7.2. A file sequence of five files opened in a paged sound
window, with page size of 30 seconds. The entire file sequence is about
four minutes long. For a single file opened in a paged window, the
current file name and the file sequence number are not displayed.
The paging scrollbar
The page navigation panel contains a paging scrollbar whose length
corresponds to the entire signal, including those portions of the signal that
are not presently in memory. The scroll box in the paging scrollbar
represents the data in the current page (i.e., the data currently in memory).
The length of this scroll box, relative to the length of the entire scrollbar,
indicates what proportion of the entire signal is included in the current
page. The position of the scroll box within the scrollbar indicates the
position (in time) of the current page within the entire signal. The start
time of the current page is displayed below the scrollbar. You can move
the page to any point within the signal by dragging the scroll box. As you
189
drag the scroll box, the displayed start time of the page is continually
updated to correspond to the position of the scroll box. If the signal is a file
sequence, the name and sequence number of the file containing the start of
the page are also continually updated as you move the scroll box. When
you release the scroll box at a new start time, Raven loads the data for the
new page position as needed from the file or file sequence and updates the
views displayed in the window. Depending on the page size and the
number and type of views shown, there may be a slight delay while Raven
updates the views.
Clicking in the scrollbar to the right or left of the scroll box pages forward
or back, equivalent to clicking the Page forward or Page back button (see
below).
Don’t confuse the functions of the two horizontal scrollbars in
a paged sound window. The scrollbar in the page navigation
panel at the top of a paged sound window refers to the time
dimension of the entire signal. The scrollbar and navigation
controls at the bottom of the window refer to the horizontal
axis of the active view for the current page.
Moving the page incrementally: the Page and Step buttons
The Page Forward, Page Back, Step Forward, and Step Back buttons
(Figure 7.2) move the current page to a new location within the entire
signal by large or small increments. The amount that the page moves
forward or back through the signal depends on the values that were
entered in the Page increment and Step increment fields of the Configure
New Sound dialog box (see “Configuring a new paged sound window”
on page 188). By default, paging forward or back moves the page by 90%
of the page size, resulting in a 10% overlap between successive pages.
Stepping forward or back moves the page by 10% of the page size by
default, resulting in a 90% overlap between successive pages.
Jumping directly to a particular time within a signal
To jump to a particular page start time within the entire signal, enter a
time value in the Page Start Time field in the page navigation panel
(Figure 7.2), and press <Enter>. You can enter a time either in units of
seconds, or minutes and seconds, or hours, minutes, and seconds. The
latter two formats use colons to separate the units, for example 1:02:03
to represent 1 hour, 2 minutes, and 3 seconds.
The current file indicator; jumping to a particular file in a sequence
For file sequences, the page navigation panel displays the name of the file
containing the start of the current page. To move the start of the page to
the beginning of a particular file, choose the name of that file from the file
name drop-down menu.
190
Timed auto-page advance
To facilitate paging through long files, you can choose to automatically
page through a file at a specified rate. The buttons to the right of the
current file indicator allow you to start and stop auto-paging as necessary.
You can also choose the amount of time that is spent on each page before
advancing by changing the value in the timed auto-page delay field next
to the start and stop buttons. You can choose whether or not to play each
page as you advance using the auto-play toggle button. If the toggle
button is on, Raven plays each page before waiting the specified amount
of time and paging forward.
Jumping to recently viewed pages
You can also navigate to recently viewed pages by using the buttons at the
bottom right of the page navigation panel. This may be useful if you are
moving around in a file to browse selections or other events.
Selections You can create a selection that spans boundaries between pages by shiftspanning pages clicking. Click once at the start or end of the interval you wish to select,
then page forward or back as needed, and shift-click at the other limit of
the interval. Once the selection is created, you can copy or save it as you
would with any selection in a non-paged window.
Reconfiguring You can change the page size and the page and step increments for a
paging for an paged sound window at any time by choosing Configure Paging... from
existing window the contextual menu for the page navigation panel (right-click with a 2button mouse or <Control> + click on a Mac). You can then enter new
values for any of the page configuration parameters in the Configure
Paging dialog box (Figure 7.3). If you adjust the page size or number of
lines while zoomed to show an entire page of sound, Raven will
automatically resize the view to show the entirety of the new page size.
However, if you are zoomed in to a shorter timespan, the view will not be
resized. If you would rather automatically zoom to view an entire page
when the page size or number of lines is changed, you can change the
corresponding entry in the preferences file from false to true.
raven.ui.soundWindow.pageSizeChanged.zoomToAllX=false
raven.ui.soundWindow.numberOfLinesChanged.zoomToAllX=false
191
Figure 7.3. Configure paging dialog
Figure 7.3. The Configure Paging dialog box, which is used to
reconfigure paging for an existing paged sound window.
Batch Operations
Raven has a number of features for working with many files
simultaneously. The batch detector allows you to perform simultaneous
detection runs on large numbers of files. The batch channel exporter
allows you to export a specified subset of channels from all the files in a
given folder. The batch correlator performs the same correlation on a large
number of files and displays the peak correlation values in a table for easy
comparison. Finally, it is also possible to filter or amplify a set of files by
using the batch filter and amplify tools. For complete descriptions of these
features, see “Batch channel export” in Chapter 8 (page 212), “Batch
correlation” in Chapter 9 (page 240), “Batch Filtering and Amplifying” in
Chapter 6 (page 150), or “Batch detection” in Chapter 10 (page 262).
Background Tasks
Progress Manager When you work with large datasets, some operations can take a long time
to run. Raven allows you to run many of these operations in the background while you continue to work. The Progress Manager keeps track of
these background tasks, showing you how close they are to completion,
how much time has elapsed since they began, and other information specific to the tasks (Figure 7.4). The Progress Manager will appear automatically when a new background task begins. If you close or hide it to work
on another task and then would like to open it again, click on Window >
Progress Manager or click the Progress Manager button in the toolbar.
192
Figure 7.4
Figure 7.4. The Progress Manager with two finished task panels visible. Note the Clean Up button in the lower left.
A new panel opens in the Progress Manager for each background task you
begin. By default, these panels will remain in the Progress Manager after
their tasks are complete, so that you can review information about the
tasks. If you would like a panel to close automatically after its task finishes, check the Close When Done checkbox in that panel. Once you
check that checkbox in one panel, subsequent panels will have the checkbox checked by default. At any point you can close a finished panel by
clicking its Close button, or clear all finished panels from the Progress
Manager by clicking on the Clean Up button.
Tasks that appear in the Progress Manager include batch correlation, batch
detection, and batch amplification and filtering (see “Batch Operations” in
Chapter 7 (page 192)). Other tasks, such as opening large files or generat-
193
ing spectrograms, do not run in the background. These tasks’ progress
bars will appear in separate windows.
By default, Raven displays a progress dialog if it estimates that a task will
take longer than two seconds. If you’d like Raven to wait a longer or
shorter period of time, change the following preference:
raven.ui.progressDialog.thresholdTime=2
Change the number 2 to the length of time you’d like Raven to wait, in seconds. For more information about how to use preferences, see “About
Raven preferences” in Chapter 11 (page 299).
Background task The raven icon in the lower right corner of the Raven window indicates
indicator whether a background task is running. Most of the time, when no
background tasks are running, the raven is perched. However, if a task
such as a batch operation or large detector is running in the background,
the raven will be in flight (See Figure 7.5).
Figure 7.5background task indicator!
Figure 7.5. Raven icon shown (a) with no background task running (b)
while running a background task.
Opening file sequences
Raven allows you to open multiple sound files as a file sequence, provided
that the files have the same file format (e.g., AIFF or WAVE), sample rate,
and number of channels. When files are opened as a sequence, the audio
data in the files are concatenated into a single sound signal that opens in
one window. A sound window containing data from a file sequence looks
and behaves the same as a single-file sound window with a few small
exceptions. When a file sequence is opened, the title bar of the sound
window displays the name of the first file followed by “etc”. To find the
name of the file that is currently displayed, position the mouse cursor over
any part of the sound view. That section’s file name will be displayed in
the mouse measurement field, just below the Raven desktop (Figure 7.6).
Also, the Cut, Delete, and Paste operations cannot be used on a file
sequence, as changing the length of the file sequence would result in the
file boundaries appearing in the wrong places.
You can open a sound file sequence either by giving Raven the name of a
list file, or by manually choosing a set of files to be opened as a sequence.
194
Using a list file A list file is a plain text file (filename extension.txt) that lists the names
of a series of sound files to be opened together as a sound file sequence. A
list file is created automatically by Raven when you record to a file
sequence (see “Recording to a file sequence” in Chapter 4, page 92). You
can also create a list file manually in a word processing or text editing
program.
To open a file sequence using a list file, choose File > Open Sound Files...
or type <Ctrl-O> (Windows, Linux) or <Command-O> (Mac OS). In the
Open Sound Files dialog box that appears (Figure 1.2, page 3), choose or
type the name of the list file, then click OK. In the Configure New Sound
Window dialog box, you can then choose either to open the entire file
sequence or to open the sequence in a paged sound window. Paged sound
windows are discussed in “Configuring a new paged sound window” on
page 188. A new sound window will appear displaying all of the audio
data from all of the sound files specified in the list file, or data from the
first page of the sequence, in the order that the file names appear in the list
file.
Format of a list file
A list file consists of a series of file name entries, one per line. Each entry
can be a simple file name, a relative path name, or an absolute path name.
If the entry is a simple file name, Raven looks for the file in the same
directory as the list file. A relative path name can start with the name of a
subdirectory within the list file’s directory, or it can use the double-dot
(..) notation to indicate the parent directory of the list file’s directory. For
example, consider a list file that resides in the directory c:\SoundData
and contains the following four entries:
firstFile.wav
moreData\secondFile.wav
..\otherSounds\thirdFile.wav
c:\frogSounds\fourthFile.wav
When opening this file sequence, Raven will look for the following files:
195
c:\SoundData\firstFile.wav
c:\SoundData\moreData\secondFile.wav
c:\otherSounds\thirdFile.wav
c:\frogSounds\fourthFile.wav
When specifying directories in list files, the forward slash (/)
and back slash (\) characters may be used interchangeably.
On a Mac, consider the list file /Applications/RavenPro1.4/Sounds/
MyList.txt that contains the following entries:
sound.wav
../examples/CanyonWren.wav
birds/Chickadee.wav
/users/Amanda/Sounds/rugbyGrunt.wav
When opening this file sequence, Raven will look for the following files:
/Applications/RavenPro1.4/Sounds/sound.wav
/Applications/RavenPro1.4/examples/CanyonWren.wav
/Applications/RavenPro1.4/Sounds/birds/Chickadee.wav
/Applications/RavenPro1.4/Sounds/users/Amanda/Sounds/
rugbyGrunt.wav
Selecting files in a You can manually specify a set of sound files to open as a sound file
sequence sequence by choosing multiple files from within the Open Sound Files
individually dialog box. There are three ways to choose multiple files to open as a
sequence:
•Click on the name of the first file you want to open in the list of files.
Then select additional files to open by clicking on their names while
holding down the Ctrl key. The names of all of the files you clicked will
be highlighted, and will be shown in the File Name field. Each file name
in the File Name field will be enclosed in double quotation marks
(Figure 7.7).
•If the names of the files you want to open are listed consecutively (with
no intervening file names) in the file list, click on the name of the first
file to open, then shift-click on the name of the last file. The names of the
two files you clicked and all files between those two will be highlighted,
and will be shown in the File Name field. Each file name in the File
Name field will be enclosed in double quotation marks.
•You can type the names of the files that you want to open in the File
Names field, with each name enclosed in double quotation marks (like
“myFile.wav”), and a space between successive quoted file names.
After you click OK in the Open Sound Files dialog box, the Configure New
Sound Window dialog box appears (Figure 7.8). In the Multiple Files pane
196
at the bottom of the Window tab, click on Open as file sequence in one
window.
After choosing either to open the file sequence in its entirety in a single
window or in a paged sound window (see “Configuring a new paged
sound window” on page 188), click OK. A new sound window will appear
displaying all of the audio data from all of the sound files specified in the
Open Sound Files dialog box, or data from the first page of the file
sequence. When Raven opens a sequence of files that were individually
specified (as opposed to given in a list file), the order in which data from
the different files are displayed is determined by the lexicographic order
of the file names.1
When data from more than one file in a sequence are visible in a view,
Raven displays dashed green file boundary markers (Figure 7.6) between the
data from each file.
Figure 7.6. Sequence of 2 files
Figure 7.6. The files “ChestnutSidedWarbler.aif” and “LarkSparrow.aif”
from the Raven Examples folder opened as a file sequence. The
dashed green vertical line shows the boundary between the two sound
files.
1. In lexicographic ordering, numbers precede letters. For example, if you were
to open two files named abc.wav and ab1.wav as a sequence, the data
from ab1.wav would be displayed first followed by the data from abc.wav.
197
Opening multiple sound files in separate windows
You can open more than one sound file at a time, with each sound
appearing in a separate window, by choosing multiple files from within
the Open Sound Files dialog. To display the Open Sound Files dialog,
choose File > Open Sound Files.... There are three ways to choose multiple
files to open:
•Click on the name of the first file you want to open in the list of files.
Then select additional files to open by clicking on their names while
holding down the Ctrl key. The names of all of the files you clicked will
be highlighted, and will be shown in the File Name field. Each file name
in the File Name field will be enclosed in double quotation marks
(Figure 7.7).
•If the names of the files you want to open are listed consecutively (with
no intervening file names) in the file list, click on the name of the first
file to open, then shift-click on the name of the last file. The names of the
two files you clicked and all files between those two will be highlighted,
and will be shown in the File Name field. Each file name in the File
Name field will be enclosed in double quotation marks.
•You can type the names of the files that you want to open in the File
Names field, with each name enclosed in double quotation marks (like
“myFile.wav”), and a space between successive quoted file names.
•You can click on a .txt file that contains a list of file names to open, like
those produced by the Raven file sequence recorder. See a description of
list files in “Using a list file” on page 195.
Once you have chosen all of the files you want to open, click OK.
198
Figure 7.7. Open dialog 2 files selected
Figure 7.7. The Open Sound Files dialog box, with two files selected by
control-clicking in the file list.
After you click OK in the Open Sound Files dialog box, the Configure New
Sound Window dialog box appears (Figure 7.8). In the Multiple Files pane
at the bottom of the Window tab, click on Open in separate windows.
After choosing either to open each file in its entirety in a single window or
in a paged sound window (see “Configuring a new paged sound
window” on page 188), click OK. Each of the chosen files will open in a
separate window, using the window preset that was specified in the
dialog.
199
Figure 7.8. Config multi-file window
Figure 7.8. The Configure New Sound Window dialog for opening
multiple files.
200
Figure 7.9 separate windows
Figure 7.9. The sound files selected in Figure 7.7 shown in separate
sound windows after selecting “Open in separate windows” in the
Configure New Sounds Window dialog.
Correcting sounds acquired at altered speeds
It is sometimes necessary or useful to acquire sounds from recordings
played at speeds slower or faster than the speed at which they were
recorded, as discussed in “Advanced options: acquiring from recordings at
altered input speeds” in Chapter 4 (page 99). Sounds that were acquired
with the appropriate speed correction in Raven 1.2 and subsequent
versions, as discussed in that section, require no further corrections.
However, if you have sound files that were acquired at altered input
speeds with older versions of Raven or with another program that did not
correct for the altered input speed, you can correct the speed when
opening the file.
Correcting speed The Speed tab in the Configure New Sound Window dialog box (Figure
when opening file 7.10) allows you to specify the factor by which the sound was slowed
down or sped up (relative to normal speed) when it was acquired. The
Configure New Sound Window dialog box is displayed after you choose
one or more files to open in the Open Sound Files dialog box.
201
Figure 7.10. Configure window-- speed tab
Figure 7.10. The Speed tab in the Configure New Sound Window
dialog box.
If you provide the appropriate speed correction factor when opening the
file, then Raven will display time and frequency axes of views, and
measurements so that they refer to the original sound at normal speed.
Sounds that are acquired at an altered speed using the
appropriate speed correction factor behave in Raven as if
they had been acquired at normal speed. Thus, to hear
sounds that were originally above or below the human hearing range, you will need to specify an appropriately low or
high speed for playback, as described in “Playback rate” in
Raven corrects a file’s speed by altering its sample rate. If you save a file
after applying a speed correction, it is saved with the new corrected
202
sample rate. The next time the file is opened (either by Raven or any other
program), its speed will be correct without any adjustment.
In order to save a file with a corrected speed, you must open
a single file (not a sequence) in its entirety (not in a paged
sound window). For file sequences, you must either apply the
correction each time you open the sequence, or else open
and save each individual file once with a speed correction.
For files that are too large to open in unpaged windows, you
must apply the speed correction each time you open the file.
203
204
Chapter 8
Multi-channel Sounds
About this chapter
Raven allows you to work with sounds containing any number of
channels. The sound cards supplied with most computers allow you to
record and play back sounds containing either one or two channels (i.e.,
mono or stereo).
To record or play back sounds with more than two channels, you need to
have a multi-channel sound card or external device installed on your
computer.
Extensible multi- Raven 1.4 includes support for multi-channel recording using both NIchannel audio input DAQ (National Instruments Data Acquisition hardware) and ASIO
device support (Audio Stream In/Out) audio input devices. The NI-DAQ technology
provides input capabilities of up to 32 channels with an adjustable voltage
range. ASIO devices provide 24-bit audio sample size capabilities. Both of
these are Windows platform devices. Raven also supports Java Sound for
Windows, which provides mono and stereo recording capabilities. The
Macintosh version of Raven supports Java Sound and also Apple's Core
Audio, which provides multi-channel recording capabilities with 24-bit
audio sample size (see Chapter 4, “Signal Acquisition (Recording)” for
more information).
In this chapter you will learn:
•the basics of working with multi-channel files
•how to export subsets of channels from a single file or batch of files
•how to create selections in multi-channel files
•how to use multi-channel files for bearing analysis
Working with multi-channel files
Opening a multi- When you open a sound file in Raven, the Configure New Sound Window
channel file dialog appears. The Channels tab of that dialog allows you to name the
sound’s channels and select a subset of them to open (Figure 8.1).
The Channels tab contains a list of all of the file’s channels, with corresponding checkboxes. Check the checkboxes of all the channels that you
want to work with. The other channels will not be opened.
205
Chapter 8: Multi-channel Sounds
You can also give each channel a name by typing it in the box next to that
channel’s number. By default, each channel’s name is the same as its number, but you may want to rename the channels based on, for example, the
locations of the microphones that recorded them or the signals you hope
to find in them. The names that you give the channels will appear everywhere in Raven that the channels are listed.
Figure 8.1
Figure 8.1. The Channels tab of the Configure New Sound Window
dialog. The channels have been given names. Channels 1, 2, and 4
will be opened, and channel 3 will not.
Displaying and The Layout tab of Raven’s side panel contains a Channels pane that lists all
hiding channels of the channels of the active sound window (Figure 8.2). Each channel is
identified by its name, or by a number if it does not have a name. A
checkbox next to each channel name controls whether that channel is
visible in the sound window. To display or hide a channel, turn the
checkbox on or off by clicking on it. You can display or hide all channels in
a sound at once by clicking the Show All or Hide All icon above the
checkboxes (Figure 8.2).
206
Figure 8.2. Channels pane in side panel
Figure 8.2. The Channels pane on the Layout tab of Raven’s side
panel.
In the Layout tab of Raven's side panel, checking Line Titles in the
Components menu causes channel names to be displayed along with the
view type and the number of each view.
In multi-channel signals, hiding or showing particular channels of a view will also hide or show those channels in other
views that are linked by Channel Visibility. (See “Linking
and unlinking views” in Chapter 3 (page 62))
Editing multi- Each channel of a multi-channel sound must be exactly the same length
channel sounds (i.e., must contain the same number of samples) as every other channel in
the sound. Therefore, editing operations that change the length of a
sound— Cut, Paste, and Delete— always apply to all channels in a sound.
Editing operations that do not change the length of a sound— Copy, Filter
and Amplify— can apply to some channels and not to others.
If a sound window contains any data, you can paste into it only if the
number of channels in the window is the same as the number of channels
of data that were placed on the clipboard by a Copy or Cut operation. If a
sound window is empty (either a new window or one from which all data
have been deleted), its number of channels is undefined, and you can
paste any number of channels into it.
Selecting channels for editing operations
By default, all editing operations apply to all channels. If you want an
operation (such as copying) to apply only to some channels, you must first
select those channels for editing. In the following descriptions, the
<Command> (Mac OS) key functions in the same way as the <Ctrl>
(Windows, Linux) key.
207
To select a channel for editing, click the channel name(s) (to the right of the
check box in the side panel’s channel panel) of the channel(s) that you
want to save; the channel name(s) will become highlighted in yellow.
<Ctrl> + click (or <Command> + click) can be used to select more than one
channel.
A channel can also be selected by holding down the <Ctrl> (or
<Command>) key and clicking the view selection button (see Figure 3.2,
page 47). To select additional channels, <Ctrl> + click (or <Command> +
click) in their view selection buttons.
To deselect a channel that is already selected, either click again on (or to
the right) of its channel name in the Layout panel, or <Ctrl> + click (or
<Command> + click) on its yellow view selection button.
If there are multiple channels selected, in order to deselect a
channel that is already selected from the Layout panel, you
must <Ctrl> + click on or to the right of its channel name.
Moving channels to display in different order
By default, the channels will be listed, top to bottom, in the order in which
they were opened. However, you can choose to rearrange the order in
which the channels are displayed. To do this, select the channel you wish
to move and then click to drag-and-drop the highlighted channel up or
down in the channel list. This will change the order the channels are
displayed in all views in the sound window.
If you choose to save the channels after rearranging them, the
channels will be saved in the new order.
208
Figure 8.3 Selected Channels
Figure 8.3. Shows a 4-channel sound displayed in a window (right) with
channels 1-3 selected in the Side Panel (left) as indicated by the yellow
highlights.
Saving a subset of If a sound contains any data, the number of channels in the sound cannot
channels from an be changed. To save a subset of channels from an open sound, first select
open signal the channels as described above. Then choose File > Save Selected
Channels As… . You can choose to Save the new file in a number of
different formats: 8-bit, 16-bit, or 24-bit .aiff or 8-bit, 16-bit, or 24-bit .wav.
Choose the file type, then click Save.
In contrast to Raven 1.2, Raven 1.4 allows an entire channel or channels of
data to be saved without a selection having to be made, and so this feature
replaces the need to select an entire view in order to save a channel.
If you want to select an entire view for saving or editing,
make sure you are zoomed to the entire signal, then choose
Edit > Select All. (This will select all of the currently viewed
data in the active view.)
209
Figure 8.4. Channel selection
Figure 8.4. A multi-channel sound window, with Channels 1 and 3
selected for editing. The view selection buttons for the selected
channels and their entries in the side panel’s Channels list are yellow.
Playing multi- For multi-channel files, Raven allows you to specify which channels
channel files should be played back. When a sound is opened, Raven defaults to
playing the first channel in all playback channels. To choose which
channels you would like to hear, you can select the appropriate channels
in the Playback tab of the side panel, or opt to only play channels
containing the active selection (See Figure 8.5). The playback tab displays
a list of the available channels, and allows you to select which channel to
play in each output. Note that you can choose not to play sound in a given
output channel by selecting “None” in the list of available channels.
210
Figure 8.5The playback tab of the side panel
Figure 8.5. The playback tab of the side panel, set to play channel 1 in
the left ear, and channel 2 in the right ear. The corresponding playback
is illustrated in the sound.
When playing an active selection, you can use the Only Play Channels
with Active Selection toggle button in the play toolbar to override the
settings in the playback tab (See Figure 8.6). If the toggle button is selected,
Raven will only play the channels containing the active selection. In the
case of a single channel selection, the corresponding channel is played in
all playback channels. When the active selection exists in more channels
than are available for playback, the lower numbered channels are played.
Figure 8.6Play toolbar toggle button
Figure 8.6. The Play toolbar, with the Only play Channels with Active
Selection toggle button selected.
211
Exporting channels
Saving selected You can export subsets of channels from an open sound file by selecting
channels the desired channels and choosing File > Save Selected Channels As...
from the menu. More information on this process can be found in “Saving
a subset of channels from an open signal” on page 209.
Batch channel To export a specific subset of channels from a collection of sound files, put
export the collection of files (or copies of the files) in one directory. Then choose
Tools > Batch > Channel Exporter... As the Input Folder, enter the name of
the directory containing the collection of files you’d like to change, then
enter the name of the Output Folder in which you want to store the new
files. Choose the channels that you would like to export and click OK
(Figure 8.7).
Figure 8.7 Batch Channel Export dialog
Figure 8.7. Batch Channel Export dialog showing Input/Output Folders,
and the channels selected to be exported (in this case, channels 1 and
3.)
Creating selections in multi-channel sounds
Selections in multi-channel sounds can be defined for any subset of the
channels in the sound (Figure 8.4). By default, when you create a new
selection in a multi-channel sound, the selection is defined only in the
channel in which you outlined the selection. To define new selections
simultaneously in more than one channel, you must first select the
channels in which you want the selections to be created. As described
above, to select channels for creating selections, hold down the <Ctrl>
(Windows, Linux) or the <Command> (MacOS) key and click in the view
selection button at the left edge of any view of each channel to be selected.
212
Figure 8.8. Channel-restricted selections
Figure 8.8. Channel-restricted selections in two channels of a fourchannel sound. Channels 1 and 3 are hidden. Selection 1 is defined in
channel 2 only; selection 2 is defined in channel 4 only; selection 3 is
defined in both channels.
When you create a selection in a selected channel, the selection will be
defined in all selected channels. When you create a selection in a nonselected channel, it will be created in that channel only.
To change the channels in which a selection is defined after it is initially
created, open the Selection tab in the side panel. Checkboxes indicate the
channels in which the active selection is defined. To change the channels
in which the active selection is defined, check or uncheck channel
checkboxes.
Bearing analysis: Beamforming, Beamogram View, and Max Bearing Measurement
It is often useful in bioacoustic analysis to use an array of sound sensors as
a means of determining the source direction of signals recorded by the
array. One common way of doing this is to process the sound with a
beamforming algorithm. Beamforming indicates only the bearing (a line
representing a specific direction) along which a sound arrives at a sensor.
The distance between source and sensor is not computed. Also, the results
of beamforming using a straight, linear array will contain right/left
ambiguities. That is, it will not be possible to obtain information regarding
on which side of the array the source is located. However, if you know the
general location of the sound source relative to the array, or if your array
is not linear, beamforming provides information that can help to
determine specific bearing locations and to potentially distinguish
different individuals that are producing sounds in the same region.
213
Figure 8.9
Figure 8.9. A visual representation of an array (the black squares along
the y-axis), a sound source (the black circle), and the bearing of the
sound relative to the origin of the graph.
Beamogram view Beamforming in Raven provides a means for determining sound source
locations, as bearing lines, relative to a 2-dimensional, multi-channel array
of sensors. The result of Raven’s beamforming process is a set of angles,
displayed in a beamogram view, that represent potential bearing angles
between a reference direction and potential bearing line directions. Raven
iteratively tests all possible locations within a range of angles specified in
the beamogram configuration. Areas of higher energy on the beamogram
indicate the relative likelihood that specific bearing angles represent
actual sound source bearings.
Beamogram slice In addition to the beamogram view, Raven also includes a beamogram
view slice view which displays the information contained in a vertical slice of
the beamogram. Hence, it displays the likelihood that a sound is coming
from each direction at a specific point in time. In the beamogram slice
view, the horizontal axis displays the range of bearing angles and the
vertical axis displays the likelihood of a signal from that direction.
214
Figure 8.10 Beamogram view
Figure 8.10. 5-channel synthesized sound, showing spectrogram,
beamogram, and beamogram slice views, with the beamogram being
the active view. The sound used to create this display is
TestBeamo_SS1500_5ChanArray_0_17_50_60_80_d143.wav and
can be found in the Raven Examples directory.
Max bearing In the beamogram, bearing angles (beams) are represented in degrees on
measurement the y-axis, and time is represented in seconds on the x-axis. To find the
bearing angle with the maximum power for a specific selection in Raven,
choosing the Max Bearing measurement will give you the angle
represented by the area of highest energy in the beamogram. See
“Measurements” in Chapter 6 (page 166) for more information on this and
other measurements.
Configuring a To create a beamogram, open a multi-channel sound file and choose View
beamogram > New > Beamogram View... or choose the New Beamogram button in the
215
toolbar. You will be need to enter information about the recording
medium, array geometry, bandpass filtering, beamforming, and channels
in each of the tabbed pages (see “Beamogram parameters” on page 216 for
details).
Choose OK to apply the configuration and close the window (alternately,
you can choose Apply to preview your parameters without closing the
window). To update parameters for an existing beamogram view, be
certain that the beamogram is the active view and choose View >
Beamogram Configuration... or right-click anywhere on the beamogram
view (or Control+click on a Mac) and choose Beamogram Configuration...
from the context menu.
Figure 8.11 Configure Beamogram dialog
Figure 8.11. The Configure Beamogram dialog, showing the different
parameter tabs (Medium, Array, etc.) across the top. Incidentally, these
are the array settings used to create the beamogram in Figure 8.10.
Beamogram When creating a beamogram view, Raven needs to know certain
parameters information about your recording setup. Information about these
parameters can be entered into the Configure Beamogram dialog, on any
or all of the tabbed pages (see Figure 8.11). More information about these
parameters is below.
Medium
Sound speed refers to the speed of sound within the recording medium,
typically either air or water. The speed of sound varies in these media
based on a number of parameters, including the temperature and density
of the media. However, a useful approximate value for the speed of sound
216
in air is 330 meters/second, and in saltwater, a useful approximate value is
1500 meters/second.
Array
On this tab, you must enter the X- and Y- locations of the sensor array
elements. The sensor positions need to be expressed using coordinates
that indicate positions (in meters) relative to one another. So, if you have
latitude and longitude values, you will need to convert these to UTM
values, and then derive the relative positions in meters.
Filtering
If you would like the signal to be bandpass filtered before the
beamforming calculations are made, enter low- and high- frequency limits
to define the bandwidth that will be passed through the filter. If your
recording unit produces a DC-offset signal, it can be useful to filter this
out for beamforming. Also, it may be useful to filter out upper harmonic
or noisy frequencies to reduce the appearance of aliasing in the
beamogram by using a bandpass filter. (As an aside, no filtering was used
in creating the beamogram in Figure 8.10.) More information on how to
filter a signal can be found in “Filtering and amplifying sounds” in
By clicking on the Advanced button in the Filtering tab, you can configure
the parameters of the bandpass filter in more detail. A bandpass filter does
not fully reject all signals outside the specified frequency range. Instead,
these frequencies are filtered to a finite level referred to as the stop band
attenuation. For the beamogram filter, this value can be set anywhere
between 10 and 200 dB. Furthermore, a bandpass filter will always have a
finite transition region between the pass band and stop band where the
signal is filtered, but not reduced to the full stop band attenuation. The
width of this region is referred to as the transition bandwidth and can be
set between 48.83 Hz and 976.56 Hz for the beamogram filter. In general,
these values can remain at their default values. However, if you are
including low frequencies in your pass band, you should ensure that the
transition bandwidth is less than the minimum band frequency.
Beamforming
The beamforming tab has several parameters to specify, which are listed
with brief descriptions below. See Figure 8.12 for a picture of the tab.
•Time Grid: Hop size, Record size
The system computes one group of samples (with length specified in
Record Size section) at a time. Before computing each subsequent
group of samples, it ‘hops’ by the number of samples or length of time
given in the Hop Size box. This process is repeated iteratively for all
beams (bearing angles). A smaller hop size will create more ‘columns’
in the beamogram.
217
Record size is the number of samples of sound over which a beam will
be computed.
It can be helpful to experiment with turning on and off the
View > Smooth Beamogram setting as a way to explore how
the Hop size and Record size settings affect the beamogram
display.
•Normalize channels
If the gain imbalance between the recorded channels is large, it may
become the main information used in the beamformer computation.
Therefore, unless the microphones and recording unit are carefully
calibrated to produce a uniform gain, normalizing the signal will
probably be useful.
•Bearing Grid: Minimum bearing, Maximum bearing, Number of
bearings
The Minimum bearing and Maximum bearing values (in degrees)
define the range over which the series of individual beams (bearings)
will be calculated. It is generally useful to restrict the range to the
directional region most likely to contain the source.
The Number of bearings is the number of beams that will be
computed. A larger number of beams will usually produce a more
accurate measurement. The resolution of the beamformer system is
defined by the difference between the Maximum and Minimum
bearings, divided by the Number of bearings.
•Bearing Coordinate System: Reference bearing, Bearings increase
When the Reference bearing is set to zero degrees, the array is set to be
perpendicular to a line pointing toward the source. However, this may
or may not be the physical case in every situation. If the array is not
“straight” the beamogram coordinates will be in line with the array
coordinates for the actual array. Using information about the possible
location of the sound source relative to the array can provide a more
accurate beamforming result. Altering the Reference bearing angle
alters the orientation of the array relative to the sensor for the
purposes of the beamformer calculation.
The direction of rotation of the bearing grid that results from altering
the reference angle can be specified as either clockwise or
counterclockwise in the Bearings increase box.
218
Figure 8.12 Beamforming tab on configure dialog
Figure 8.12. Parameters contained on the Beamforming tab in the
Configure Beamogram dialog. These parameters were used to create
the beamogram shown in Figure 8.10.
View channels
You can create different subsets of channels to use with the beamformer
calculation and compare the results. Depending on the quality of signals
received by each sensor, it may be useful to view different combinations of
sensor signals in the Beamogram view. Each combination is called a
Channel. The limit on the number of different combinations that can be
viewed simultaneously is equal to the number of channels in the sound.
Channels are added to the end of the list and are removed from the end of
the list.
Hints for working A useful way to work, after setting the initial parameters, is to ‘steer’ the
with beamograms beamformer by adjusting the reference bearing and the min and max
bearings. Begin with a large number of bearings (200 or more) to create a
high resolution. As mentioned before, filtering the signal can also be
helpful. Figure 8.13 gives a graphical representation of how orienting the
beamformer can work.
219
Figure 8.13 Beamformer orientation
Figure 8.13. Orienting the beamformer using the Maximum, Minimum,
and Reference bearing parameters.
220
Chapter 9
Correlation
About this chapter
This chapter explains how Raven’s correlation tool works. Correlations are
a way to perform quantitative comparisons between spectrograms or
waveforms. In this chapter you'll learn:
•some general information about correlation
•how correlation functions are calculated
•how to use the correlation function and specify options common to both
spectrogram and waveform correlations
•about issues and options specific to each type of correlation
•about batch correlation
Overview of Correlation
Correlations are performed by 'sliding' two inputs (either two
spectrograms or two waveforms) past each other in time. At each time
offset, a correlation value between the inputs is calculated. These correlation
values are then plotted versus time in a correlation plot to show a measure
of similarity between the inputs. The time axis of the plot is shown relative
to the first input, indicating how far the second input has been offset, so a
peak at a positive lag, or time offset, indicates that the second signal
occurs at an earlier time than the first (see Figure 9.2.).
Correlation types Raven provides the capability to run correlations between either two
spectrograms or two waveforms.
•Correlating spectrograms: Usually when correlating spectrograms, the
most important information provided is the peak correlation value
which shows the similarity between the spectrogram images.
•Correlating waveforms: Waveform correlation can help determine the
lag at which two signals most closely resemble each other, as shown in
Figure 9.2.
Calculating Correlation Functions
Spectrogram For each lag Δt, Raven calculates either a normalized or non-normalized
correlations correlation value CΔt between two spectrograms (For more on the effect of
normalization, see “Normalization” on page 226). If the Normalize option
is selected, each correlation value is calculated using the following:
221
Chapter 9: Correlation
n FFT
  ( Xt, f ⋅ Yt + Δt, f )
Formula 9.1
t = 1f = 1
C Δ t = -------------------------------------------------------------------------------------------n FFT
n FFT



2 
2

( X t, f )
( Y t, f )



t = 1 f = 1
 t = 1 f = 1



where n equals (N1+N2) - 1 and N1 and N2 are the numbers of frames in the
two spectrograms (for a discussion of these variables, see “Spectral
analysis of time-varying signals: spectrograms and STFT analysis” on
page 331). Note that this formula corresponds to a correlation using a
biased rather than an unbiased normalization (For more on the distinction
between biased and unbiased normalization, see “Biased/Unbiased” on
page 227). FFT equals the number of frequency bins, which must be the
same for the two spectrograms being correlated. Xt,f and Yt+Δt,f are the
amplitude values (in this case, the power of the spectrogram) in the two
spectrograms at frequency f and times t and t+Δt, respectively. The
normalized correlation value for spectrograms can vary between 0 and 1.
A correlation of 0 means that the non-zero values in the two spectrograms
do not coincide at all; a correlation of 1 indicates that the two signals are
identical (given the lag Δt.) Successive correlation values are calculated by
incrementing the value of Δt in steps equal to the time grid resolution
(which must be the same for both spectrograms) in effect sliding the two
spectrograms past each other in time. If normalization is turned off, only
the numerator of Formula 9.1 is used.
222
Figure 9.1
Figure 9.1. A spectrogram correlation between two calls of a Blackcapped Vireo.
Waveform For each lag, Raven calculates either a normalized or non-normalized
correlations correlation value between two waveforms. If the Normalize option is
selected, each correlation value is calculated using:
n
 ( xt ⋅ yt + Δt )
Formula 9.2
t=1
C Δt = -------------------------------------------------n
n



2 
2

yt
xt



t = 1  t = 1 


where n equals (N1+N2) - 1 and N1 and N2 are the numbers of digitized
samples in the two waveforms. xt and yt+Δt are the values of sample
numbers t and t+Δt of the two waveforms, respectively. If the two signals
differ in length, the shorter signal is zero-padded at the end to the length
of the longer signal. The correlation value for waveforms can vary
between -1 and 1 (If you choose to plot the complex envelope of a waveform
correlation, the values that are plotted vary between 0 and 1, as discussed
in “Waveform correlations” on page 237).
223
A correlation of 0 means that the signals are orthogonal1; a correlation of 1
indicates that the two signals are identical; a correlation of -1 indicates that
the signals are identical in magnitude, but opposite in phase.
Successive correlation values are calculated by incrementing the value of
Δt in steps equal to the inverse of the sampling frequency, in effect sliding
the two waveforms past each other in time. If normalization is turned off,
only the numerator of Formula 9.2 is used.
Although Formula 9.1 and Formula 9.2 are written in the
time domain, Raven actually performs these computations in
the frequency domain. This improves the performance of the
correlator.
Figure 9.2
Figure 9.2. A normalized waveform correlation plot. The peak
correlation lag is shown on the x-axis by the vertical magenta line. Note
that a positive correlation lag indicates that the first waveform is ahead
of the second in time.
1. Whether or not two signals are orthogonal depends on their frequency content and on their relative phase. For example, sinusoidal signals of different
frequencies are orthogonal, as are signals of the same frequency that are 90o
out of phase with each other.
224
Using the correlation tool
To perform a correlation, choose Tools > Correlator and select the two files
to be compared. Next, choose whether to correlate the waveform or the
spectrogram views and select any appropriate parameters (see below for
more information on specific options).
Note that both signal files must be recorded at the same sample rate in order to be correlated.
Figure 9.3 Correlation Dialog Box
Figure 9.3. The Correlation dialog box. After selecting which two files to
correlate, you can choose which type of view to correlate (waveform or
spectrogram) along with other parameters and options as well.
Peak correlation values can be important when comparing spectrograms;
however, it is important to note that there are serious limitations to
correlation comparisons. Spectrogram correlation is not a tool for
generalized “pattern recognition” and it is important to remember that the
measured similarities are simple and narrowly defined and may not be
relevant in a given context. (see “Spectrogram correlation” on page 229)
225
The timing of a peak correlation value from a waveform correlation plot is
often useful as well. It can determine the lag at which two inputs (possibly
from the same source signal) most closely match each other. Usually,
waveform correlations are less useful than spectrogram correlations for
assessing “similarity” between signals in an intuitive way. This is partly
because (unlike spectrogram correlations) waveform correlations are
sensitive to phase differences in inputs that our auditory system does not
detect.
Canary users will find much of the correlation functionality similar,
although there are a few added features.
Band Pass Filter Choosing to add a Band pass filter is helpful for signals containing noise.
Checking the filter box allows you to enter a lower and upper frequency
limit that will filter the data before it is correlated.
In most applications, filtering is advisable for both waveform and
spectrogram correlations. By selecting a frequency band corresponding to
the relevant signal(s), you reduce the effect of any other noise or signals on
the correlation values.
The effect of filtering on the correlation function (for both waveforms and
spectrograms) depends very much on the particular signals being
correlated. If neither file contains much energy outside the frequency band
occupied by the signals, filtered correlations may not differ much from
unfiltered correlations. If there is a significant amount of energy outside
the frequency band of interest, the difference between the filtered and
unfiltered correlations can be much larger. Also, it is important to note
that filtering may either raise or lower correlation values, depending on
the particular signals being correlated.
Normalization If Normalize is checked, the sum of the products of the data values from
the two signals is divided by the square root of the product of the sums of
values from the two signals, as indicated in Formula 9.1 and Formula 9.2.
The units in the numerator and denominator cancel and the correlation
value is scaled to a dimensionless value. For spectrograms, which contain
only non-negative amplitude values, the normalized correlation value is
always between 0 and 1. For waveforms, which can contain positive,
negative, and zero values, the normalized correlation varies between -1
and 1.
If Normalize is left unchecked, the correlation is calculated as the sum of
the products of the data values from the two signals (i.e. just the
numerator of Formula 9.1 and Formula 9.2.) A non-normalized correlation
is given in arbitrary units.
Normalizing a correlation will compare the overall 2-dimensional shape
of signals but will ignore any amplitude differences. Examples of
normalized and non-normalized correlations, and how the overall
amplitude level affects the correlation values, can be seen in Figure 9.4.
Unless your application requires that similarity measurements
226
incorporate information about the absolute amplitude levels of the signals,
you should most likely leave the Normalize option checked.
Figure 9.4 Normalizing a correlation
Figure 9.4. Comparison of normalized and non-normalized
correlations. (a) Waveform of a portion of song from a Chestnut Sided
Warbler. (b) The same waveform amplified by a factor of 2. (c)
Normalized correlation between (a) and (b). This correlation is identical
to the correlation between (a) and itself. (d) Non-normalized correlation
between (a) and itself. Note the change in axis scale from units to
gigaunits (e) Non-normalized correlation between (a) and (b). Notice
the differences in correlation value amplitudes among the plots.
Biased/Unbiased By default, a biased normalization is performed during correlations.
However, you can choose to perform an unbiased normalization simply by
227
checking the Unbiased box. The expected value of an unbiased
normalization equals the quantity it estimates; however, this sometimes
leads to large variance in endpoints because only a few data points are
used. To avoid large variations at correlation endpoints, do not check the
Unbiased box.
Subset Full normalization uses values from the entire length of both parent
Normalization signals in calculating the normalization coefficient. However, this can
result in high correlation values at locations of greater intensity in a longer
sound, even if the features of the sounds are dissimilar. When the Subset
Normalization checkbox is checked, only the overlapping portions of each
sound are used in normalization. By accounting for variations in noise
level within the longer sound, this provides a more accurate correlation
between a short segment and a longer sound. Figure 9.5. illustrates this
difference when the Black-capped Vireo recording is correlated with a
single call using both full and subset normalization.
228
Figure 9.5. Comparison of correlations performed with full and subset
normalization. (a) a Black-capped Vireo recording. (b) A short selection
taken from the recording. (c) Correlation between (a) and (b) performed
using full normalization. Notice that the peak correlation is relatively
small (about 93 mu) and occurs at the incorrect location. (d) Correlation
between (a) and (b) performed using subset normalization. The peak
correlation value is now 1, and occurs at the predicted location.
However, when performing a correlation between two short clips, subset
normalization can introduce extra peaks near the endpoints of the sounds,
when only a small portion of the sounds are used in normalization. For
example, in Figure 9.6, a high correlation value is obtained when the end
of one selection is correlated with the beginning of the other, despite the
dissimilarity of the sounds.
Figure 9.6Subset normalization short sounds
Figure 9.7. A correlation between two Black-capped Vireo calls. Using
subset normalization (left image), a peak is introduced when the end of
one call is compared with the beginning of the other. However, using full
normalization (right image) this peak is not noticeable.
Spectrogram correlation
Often, the only point of interest in a spectrogram correlation plot is the
maximum value of the correlation function. The peak correlation value
can provide a quantitative measure of one type of similarity between
spectrograms.
229
Figure 9.8 Spectrogram Correlation
Figure 9.8. The two spectrograms shown were produced in succession
during the song of a Chestnut Sided Warbler. Above in green, you can
see the correlation between the two spectrograms. The entire
correlation graph is selected using Edit > Select All, and in the selection
table, the peak correlation measurement of 0.813 is included. Note that
the peak correlation and peak lag are shown on the axes and by the
magenta lines on the correlation plot, in addition to being shown in the
selection table in their measurement columns. The spectrograms are
zoomed in frequency to show the detail.
While the peak value of a spectrogram correlation function can provide an
objective, well-defined, repeatable, and comparable measure of the
similarity of two spectrograms, it is NOT a tool for generalized “pattern
recognition”. The “similarity” that is measured is simple and narrowly
defined, and may or may not be appropriate to the research question
being asked. The usefulness of spectrogram correlations as a measure of
similarity thus depends very much on the specific context in which they
are being used. The best way to develop a feel for how to interpret
spectrogram correlations is to experiment with correlating a variety of
spectrograms.
230
Figure 9.9 High vs Low Peak Correlation values
Figure 9.9. Two correlation windows, one showing two sounds with a
low peak correlation value (left), and the other showing two sounds with
a high peak correlation value (right). Note that although the signals on
the left appear to be similar, their high energy components are offset in
frequency around the 4314 Hz line, marked by the magenta frequency
position marker in the view. Raven does not do any frequency shifting
when it performs spectrogram correlations. The peak correlation value
on the left is 9.85 milli-units, whereas the peak on the right is 0.463
units, or 463 milli-units. Always be sure to check the units on the y-axis
of the correlation view.
When Raven displays a correlation plot, it allows you to scroll the time
position of the view to see how the two parent signals line up at different
points in the correlation plot. The default view when a correlation plot is
initially shown is to show the lag at the peak correlation value, and to have
the parent signals lined up at their peak correlation. Moving the time
scroll bar in the view allows you to see how the parent signals line up at
other correlation values, as shown in Figure 9.10. for spectrogram
correlation. The same type of time scrolling can be done with a waveform
correlation.
After running a spectrogram or waveform correlation, you can choose to
playback either of the parent sounds directly from the correlation
window. When a sound view is active, the playback features are available
231
and apply to the active sound. However, the playback features are
disabled in the correlation view.
Figure 9.10. Three correlation windows of the same correlation plot
scrolled to three different time positions. The first shows a lesser peak
to the left of the highest peak, the second shows the main peak, or
peak lag, the time position at which the signals are most highly
correlated, and the third shows a lesser peak to the right of the highest
peak. Observe how in both the first and the third, two of the three calls
in each signal are aligned, but that the correlation value is much higher
when all three are aligned in time.
Scale of Raven can compute spectrogram correlations using either the logarithmic
spectrogram power power values from the spectrogram, in units of decibels, or using the
values linear power values, in units of squared amplitude units. Linear power
values are the traditional method used to compute spectrogram
correlations and provide a good spread between spectrograms that look
similar and those that look different. Because the logarithmic scale
compresses the range in which power values exist, the corresponding
correlation plots also tend to be more concentrated with higher peaks,
which can lead users to think that signals are similar when they are really
dissimilar. However, it is often possible to minimize this effect by
adjusting the spectrogram clipping parameters of demeaning spectrogram
values before correlation. (For more information on spectrogram clipping
and demeaning see “Demeaning of spectrogram values” on page 233 and
“Spectrogram correlation parameters” on page 234). Also, since
logarithmic power values are used to display spectrograms and make
measurements in Raven, using logarithmic values in correlations better
coincides with these tools.
Figure 9.11 shows two normalized spectrogram correlation plots using
linear and logarithmic power values. Notice that using logarithmic as
opposed to linear power values changes the maximum correlation value
and location, as well as the shape of the correlation plot. The peak
correlation values are marked on the y-axis by the horizontal magenta line
232
and the spectrograms are shifted in time to show the best alignment of the
images, as indicated by the peak correlation lag, which is the vertical
magenta line.
Figure 9.11Linear vs logarithmic values
Figure 9.11. A comparison between a correlation using linear power
values (on the left) and the same correlation using logarithmic power
values (on the right).
Demeaning of When performing a spectrogram correlation, you can opt to demean the
spectrogram values spectrogram values before correlating the files. When Demean
Spectrogram Values is checked, Raven subtracts the average value from
each value in a given spectrogram before applying the correlation
formula. This can be useful for comparing sounds with very different
average power values or to expand the range of logarithmic power values.
Unlike standard correlation, demeaning uses negative correlation values
for sounds that are very dissimilar.
For example, Figure 9.12 shows the correlation between two dissimilar
Black-capped Vireo selections, performed with both linear and
logarithmic power values, with and without demeaning. When using
linear power values, demeaning changes the relative heights and shapes
of the peaks only slightly, but the effect is much more apparent for
logarithmic power values. Without demeaning, using logarithmic power
values results in a deceptively high correlation value and shows few of the
surrounding features. However, once the spectrograms are demeaned, the
peak correlation value is more realistic and the smaller features become
apparent.
233
Figure 9.12Spectrogram Correlation Demeaning
Figure 9.12. Spectrogram correlations for two Black-capped Vireo calls
using (a) linear values and no demeaning, (b) logarithmic values and no
demeaning, (c) linear values with spectrograms demeaned, and (d)
logarithmic values with spectrograms demeaned.
Spectrogram You can configure the view and results of spectrogram correlations (see
correlation Figure 9.15) as you would adjust display parameters for typical
parameters spectrograms. In general, increasing the overlap will decrease the hop size
and increase the number of frames in the spectrogram for a more detailed
correlation. Decreasing the overlap will oppositely affect the hop size and
number of frames. Increasing the window size will result in more
frequency bins in the DFT and a more detailed analysis of frequency, but
at the cost of time resolution. Decreasing the window size will oppositely
affect the frequency bins and time resolution.
234
Adjusting these parameters can noticeably affect the resulting
correlations. Since the correlator scans the spectrograms past each other in
time, not frequency, an increase in time resolution provides a more
detailed correlation, but reduces sensitivity to the frequency distribution
of the sounds. For example, Figure 9.13. shows a correlation between two
Black-capped Vireo calls. Notice how the top call has a band of high
energy around 3800 Hz, while the bottom call has a gap in that range.
When the correlation is performed with a large window size, this
difference is reflected in a low correlation value. However, as the window
size is decreased, the loss in frequency resolution results in a higher
correlation value as this difference is smoothed out. For more information
on configuring spectrogram display parameters, see Appendix B, “A
Biologist’s Introduction to Spectrum Analysis”.
Figure 9.13. Correlations between two Black-capped Vireo calls
performed using different window sizes. (a) 256 frames (b) 64 frames
(c) 16 frames.
Effect of
spectrogram
clipping on
correlations
When performing a correlation, you can choose to clip the individual
spectrograms below a certain level before performing the correlation. For
more on spectrogram clipping, see “Clipping level” in Chapter 5 (page
123). Because of the way correlations are calculated, the effect of
spectrogram clipping is much more significant when using logarithmic, as
opposed to linear, power values. Spectrogram clipping can reduce the
effect of noise on the correlation, but may also have undesired side effects.
For example, clipping spectrogram values to a small finite value can help
to accentuate distinct features of a signal that might otherwise be blurred
by noise. However, clipping values to no power (-infinity dB) while
performing a logarithmic correlation may return an empty correlation
view.
235
Figure 9.14 shows two autocorrelations of a Spotted Hyena call using
logarithmic power values. The correlation on the left does not clip any
spectrogram values, whereas the one on the right clips all values below 60
dB to 0 dB. By severely reducing the surrounding noise, clipping these
spectrogram values exposes the secondary peaks in the correlation that
were not previously visible.
Figure 9.14spectrogram clipping in correlations
Figure 9.14. Autocorrelation of a Spotted Hyena call performed without
spectrogram clipping (left) and with values below 60 dB clipped to 0
dB(right).
236
Figure 9.15 Spectrogram correlation options
Figure 9.15. The spectrogram configuration dialog box. You can adjust
the view of spectrogram correlations by changing these parameters.
Waveform correlations
When working with waveforms, one might be interested in the time at
which the peak correlation value occurs. To determine the position of a
sound in space, for example, an array of microphones in a known
geometry can be used to record onto separate but synchronized recording
tracks. The lags of the correlation peaks between signals on these
synchronized recordings then indicate the delays between the arrival
times of the sound at different microphones. These time delays can then be
used to calculate the location of the sound source relative to the positions
of the microphones, based on known information (including the speed of
sound and the microphone array geometry). However, in most
applications, waveform correlations are less useful than spectrogram
correlations for assessing the degree of “similarity” between signals in a
way that is intuitively satisfying (See “Using the correlation tool” on
page 225)
237
Figure 9.16 Waveform Correlation
Figure 9.16. Waveform correlation between song syllables
simultaneously recorded at two different microphones. (b) Waveform
from microphone 1. (c) Waveform from microphone 2. (a) The
correlation between (b) and (c). The time delay of -.026 s indicates that
the bird was 8.9 m closer to microphone 1 than to microphone 2
(assuming a speed of sound of 344 m/s).
Complex envelope If you are performing a correlation between two waveforms, you can
check the Complex envelope box which will display the complex envelope
of the correlation function, as opposed to the correlation function itself. A
complex envelope varies between 0 and 1. The relationship between the
complex envelope and the correlation function itself is illustrated in Figure
9.17.
238
Figure 9.17 Complex envelope
Figure 9.17. (a) The waveform correlation of two sounds. (b) The
complex envelope of the correlation function in (a). (c) The complex
envelope function superimposed over the correlation function. For
clarity of illustration, the complex envelope (in red) is shifted slightly
upward in the figure from the raw correlation plot (in green) and the
horizontal axis has been zoomed in to show more detail.
239
Correlation functions always contain high-frequency oscillations, which
are related to the frequencies present in the signals being correlated. If the
signals are approximately sinusoidal (i.e., at any moment, most of the
energy in each signal is concentrated at a single frequency, as in the
frequency-modulated whistles common in bird song), their correlation
function will itself be close to an amplitude modulated sinusoid. In this
case, the complex envelope is roughly equivalent to the amplitude
envelope of the absolute value of the correlation function. If the signals
being correlated are spectrally complex (with energy distributed over
many frequencies, as in human speech), their correlation function contains
high-frequency oscillations that are generally not sinusoidal.
Taking the complex envelope removes much of the high-frequency
oscillation in a correlation function, which can make it easier to visually
identify the peak of a waveform correlation. Since the qualitative
relationship between the appearance of the complex envelope and the raw
correlation plot depends somewhat on the signals being correlated, you
should experiment with the type of signals that you work with in order to
get a feel for the relationship between the two types of plot.
Batch correlation
Raven’s batch correlator provides a mechanism for automatically
performing the same correlation operation on an arbitrarily large number
of files and saving the results. For example, a batch spectrogram
correlation lets you perform correlations of many files at a time, using a set
of spectrogram parameters that you specify only once. The batch
correlation process runs in the foreground so you have to wait for it to
finish before you can perform any more analysis within Raven, but you
can use other applications while the batch correlator is running.
To run the batch correlator, choose Tools > Batch Correlator... from the
menu bar. This will display the batch correlation configuration dialog
Figure 9.18.
240
Figure 9.18batch correlation configuration dialog
Figure 9.18. The batch correlation configuration dialog.
Input and output For Raven’s batch correlator, you must specify which files are to be used as
input and where the output file should be saved. The input files must all
be stored in the same folder, and the output file can be saved in a folder,
which can be different from the input folder.
For batch correlation, there are two sets of file inputs (see Figure 9.18).
Each set can consist of a file or folder. If you choose a folder, then all sound
files within that folder will be used as input to the batch correlation
process.
After you select the input files or folders, you must select the output
options (see Figure 9.18). The default settings are to save the output to a
text file and to view the results within the Raven window. One of these
options must be chosen for the correlator to run. You can change the
output folder and name by using the Browse... button, or just enter the
name of the output file by typing in the text box.
241
All other options within the batch correlation configuration dialog are
identical to those in the correlation configuration dialog box, as discussed
earlier in this chapter.
Running the When you click the OK button, Raven starts calculating the correlations
correlator and displays a status window, titled “Spectrogram Correlation” or
“Waveform Correlation” depending on your selected option, as shown in
Figure 9.19. The progress bar in the status window shows the progress of
the entire batch process. Clicking the Cancel button stops the batch
correlation.
Figure 9.19 spectrogram correlation window
Figure 9.19. The correlation status window (for a batch spectrogram
correlation, in this case).
The batch correlator computes correlations for each possible pair of files in
the two input sets. When all of the correlations have been calculated,
Raven displays a window containing the correlation table (Figure 9.20).
The title of the window specifies whether the correlation was of waveform
or spectrogram views. The correlation table contains a peak correlation
value (in correlation units) for each pair of files. Clicking on the Lags radio
button will change the view to show the lag values (in seconds), or lags at
which the peak correlation values occur. If the Colors option at the top of
the window is selected, Raven will color code the peak correlation values
using the following system: Blue for low scores, then green, yellow,
orange, and red for higher scores.
To use a continuous colormap for batch correlation results as
opposed to the ten default colors, change the preferences
entry: raven.ui.correlation.colorMap.continuous from false to
true. Once this preference has been changed, you can also
change the RGB for low, medium, and high correlation values
by editing the corresponding preferences.
242
By default, the correlation table rows are sorted by the name of File 1.
However, clicking on a column heading will sort the table by that
column’s values. The columns of the correlation table may be resized by
positioning the mouse pointer on the line between the column headings,
and then dragging the line to a new position. The window itself may be
resized in order to show more columns, or the horizontal scroll bar at the
bottom of the window may be used to scroll to different columns in the
table. If there are enough rows in the table such that they cannot all be
seen, then a vertical scrollbar is provided to allow you to scroll to other
rows.
Figure 9.20 Batch Correlation Table
Figure 9.20. A batch correlation table showing peak correlation values
for each file pair using linear power values in the correlations.
To view the correlation function of a single correlation, just double-click
on any peak or lag value in the table. Raven will launch a new correlation
window showing the correlation function in one view along with the two
files that were correlated, each in their own views (see Figure 9.21). This
window is identical to the one that would be shown if a correlation had
been run on the individual files. Any of the individual sound files can also
be opened by double clicking on the row heading.
243
Figure 9.21 single correlation view from table
Figure 9.21. By clicking on a single cell in the batch correlation table
(top), Raven displays the individual correlation view between the files in
a new sound window. Note that the frequency position marker in the
correlation view marks the peak correlation value, and the time position
marker marks the peak lag value. You can also add these
measurements to the selection table, as shown in the figure.
Correlator output can also be saved to a tab-delimited text file, where the
correlation peak table and correlation lag table are displayed one after the
other (see Figure 9.22). You can specify the output file name on the initial
correlator dialog.
244
Figure 9.22 correlation table output to txt file
Figure 9.22. A correlation table saved as a file, and subsequently
opened using Microsoft Excel.
If you are correlating the contents of a folder with itself, the peak
correlation value for X*Y and Y*X will be in the same table, but one lag
will have a positive value and the other a negative value, as shown in the
lag section of the spreadsheet of Figure 9.22.
For multi-channel files correlated in the batch correlator, each channel is
displayed as its own row or column (Figure 9.23). If the same files were
correlated using Tools > Correlator, the resulting batch correlation table
would not be created, and all channel correlations would be in their own
windows. For ease of viewing the results, you should use the batch
correlator when dealing with multi-channel files.
Figure 9.23 multi-channel correlation table
Figure 9.23. A correlation table with results from a multi-channel file
correlation. In this correlation, the input files are both the same (Bird
array 4-channel.aif), so each channel of the file is correlated with itself
and with the other channels.
If a single file is batch correlated with a folder of files, the resulting table
will either be a single row (if the file is specified as input 1) or a single
column (if the file is specified as input 2). See Figure 9.24 for an example.
245
Figure 9.24 file and folder correlation
Figure 9.24. Correlations between one file and a folder of files. (a) the
file was specified as input 1 and the folder as input 2. (b) the folder was
specified as input 1 and the file as input 2.
Correlator example
This example shows, step by step, how to run a batch correlation on a set
of provided files and compares two specific correlations to demonstrate
how the correlator works.
Open sound file Shipped with Raven, you will find an audio file (located in the Examples
and selection table folder) named BlackCappedVireo.aif and a text file containing selections
(located in the Selections folder) named BlackCappedVireo.selections.txt.
Opening the sound file and then opening the selection table should give
you the following image (Figure 9.25).
246
Figure 9.25 Black-capped Vireo with selections
Figure 9.25. Recording of a Black-capped Vireo, showing 44 selections
within the recording (achieved by opening the BlackCappedVireo.aif file
along with the BlackCappedVireo.selections.txt file).
Save selections into Next, choose File > Save All Selections in Current Table As... from the
a new folder menu and click on the File Names tab. In the Directory field, type “/BCV”
after the default “Selections” entry, creating a new folder named BCV (see
Figure 9.26). Also, adjust the Sound File entry to read “BCV<ii>.aif (this
will name the selection files BCV01.aif, BCV02.aif, BCV03.aif and so on.)
Skip over the list file name and annotations box, as they are not important
for this example. For information on these options, see “Saving all
selections” in Chapter 6 (page 161). Finally, clicking the Save button will
complete the process, saving the named selections into the newly created
folder.
247
Figure 9.26 Save All Selections
Figure 9.26. The Save All Selections As dialog box with fields
completed as described in “Save selections into a new folder” on
page 247.
Run batch After saving all of the selections to the BCV folder, you can now run a
spectrogram batch correlation of all the files in the folder against each other. (Note that
correlation you could also choose to correlate a single file against all the files in the
folder. However, this example uses the BCV folder as both input 1 and
input 2.)
To perform the batch correlation, choose Tools > Batch Correlator... which
will display the appropriate dialog box. For both input files, choose
Browse... and select the new folder you created, BCV (it should be in the
Selections folder in your Raven directory). Next, adjust the name of the
output text file to read “BlackCappedVireo.specCorr.txt” and make sure
that Spectrogram (under the correlate section) and Normalize (under the
options section) are selected. After verifying your inputs with Figure 9.27,
select OK to begin the correlation.
248
Figure 9.27 Batch Spectrogram correlation dialog
Figure 9.27. The Batch Correlation Configuration dialog with
information filled in to correlate the selections contained in the BCV
folder against themselves.
The correlation After running the correlation, you should get a 44 x 44 table of correlation
table values (Figure 9.28) where each cell in the table represents the correlation
of one of the files against another of the files.
249
Figure 9.28 correlation table
Figure 9.28. A portion of the correlation table from the previous
example (correlating all files in BCV against each other).
You can view individual correlation functions by double-clicking in any
cell showing a peak value (or lag value). Double-clicking the boxes for
BCV14 * BCV01 and for BCV17 * BCV03 displays two sound windows.
Each shows the correlation function view (top) as well as the two
spectrogram views for the sounds that were correlated.
As shown in Figure 9.29, the correlation peak values and peak lags are
marked with the magenta frequency position marker and time position
marker, respectively. In each window, select the entire correlation
function, and then display the selection table. You can format the table to
include only the measurements for Peak Correlation and Peak Lag. For a
refresher on working with selection tables, see “Selection Tables” in
After doing this for both sound windows, you should have something
similar to Figure 9.29. This example demonstrates the difference between
correlation functions with high and low peak correlation values. In this
example, the correlation values seem to correspond with the similarity
between the calls being compared. While this is an important example, it
is essential to remember that not all results will be this evident; not all
250
correlation values will correspond to visual similarity/dissimilarity of
spectrograms.
Figure 9.29 high and low correlation values
Figure 9.29. (a) Linear power spectrogram correlation (top view) of
BCV14 (middle view) with BCV01 (bottom view) with a low peak
correlation value of 0.00813 (see selection table at bottom of window).
(b) Linear power spectrogram correlation (top view) of BCV17 (middle
view) with BCV03 (bottom view) with a high peak correlation value of
0.37982 (shown in selection table at bottom of window).
251
252
Chapter 10
Detection
About this chapter
When users work with large datasets, they often seek to first identify all
sounds of a particular type, then they make measurements on the selected
sounds and perhaps associate annotations with the sounds to classify
them as to type or source. The process of manually browsing through
hours, days, or months of recordings can be daunting.
An alternative to manually browsing through large datasets is to use
automatic detection. Detection is the process of finding specific sounds of
interest within recordings. Often, the signals of interest are short in
duration compared to the overall recording. Detection can be a useful tool
when compared to the alternative process of manually browsing days of
recordings in order to find small samples of interesting sounds.
Automatic detection software may be able to reduce the amount of time
that it takes to analyze a recording. Detection is accomplished through the
use of a specific detector - the mechanism which runs a detection
algorithm to detect specific signals of interest. Users must be aware of
trade-offs that they will need to make before deciding if detection is the
correct approach for their analysis.
In this chapter, you will learn about:
•the four modes of Raven detectors
•trade-offs involved with using detectors
•interactive detection
•full mode detection
•batch detection
•real-time detection
•saving detected selections
•instructions for detectors provided with Raven
The four run-modes of Raven detectors
In order to shorten the amount of time that users take to analyze their
recorded datasets, Raven has a detector infrastructure built around
Raven’s extensible software architecture. Raven provides an Application
Programming Interface (API) as well as a reference implementation of a
specific detector meant to get users started with detection. Developers
253
Chapter 10: Detection
interested in creating their own detectors should contact the Raven team
at [email protected] for more information.
Detectors in Raven can be run in one of the following four modes:
1. Interactive detection: Within a short recording or a set of pages of a
long recording, users can test several sets of parameters in order to see
how a detector will perform. Available in the View menu and in the
context menus of applicable sound views, interactive detection
searches through the paged area of a sound file to find areas that match
the detector parameters. Results from multiple interactive detectors can
be displayed simultaneously, and are color-coded for easy
discrimination, allowing the user to run and compare multiple sets of
detector parameters.
2. Full detection: Once specific parameters are chosen that work for short
sections of a recording, Raven can run a full detection process on the
entire recording even if only a single page is shown in the window.
Located in the Tools menu and through the “Run Full” button in the
interactive detector panel (see Figure 10.1), full detection is useful for
bulk processing of large datasets.
3. Batch detection: If users want to find similar signals in many different
sound files, Raven’s batch detector can be used to run a full detection
process on many sound files at once. The same detection parameters
will be used on each sound file, and Raven will automatically save a
selection table of detection results for each sound file.
4. Real-time detection: When an interactive detector is opened in a
recorder window, Raven can detect events as they occur by using a
real-time detector. Real-time detection identifies signals of interest as
they are recorded and marks them as selections. These events can then
be reviewed in the recorded file sequence or saved as clips for later
review. You can run real-time detection at the same time that you save
to file-sequences, or you can post-process the file sequence by running
a detector on the full sequence.
Figure 10.1 Run Full
Figure 10.1. The Run Full button in the detector tab on the side panel
254
Raven comes with two sample detectors; a band limited energy detector,
and an amplitude detector. These are described in more detail in “Full
detection mode” on page 261. Both of these detectors require a specific
sound view. The energy detector uses the spectrogram view, whereas the
amplitude detector uses the waveform view. For interactive detection, the
appropriate view must be active in order for the detector to be available
through the view menu. When running a detector in full detection mode
from the tools menu, Raven will prompt you to open the necessary view if
it is not already open.
Interactive detection
Running a detector First, open the sound file, making sure that the appropriate view is active
on an existing in the window. Next, open the detector by selecting View > Interactive
sound file Detectors... > [Detector Name] from the toolbar, or by opening the view’s
context menu and selecting Interactive Detectors... > [Detector Name].
If the recording is long, choose to open it one page at a time
with a page size that is large enough to capture one or more
of the signals of interest, but not so large that you run into
memory issues. For further information on memory management, go to “The Memory Manager” in Chapter 11 (page
309)
Once a detector is selected, the configuration dialog will be displayed,
allowing the user to set certain detector parameters. The parameters in the
upper half of the configuration dialog are the same for all detectors,
whereas the parameters in the lower half are specific to the individual
detector. These general parameters are described below. For information
on parameters specific to the 2 detectors included with Raven, see “The
Band Limited Energy Detector” on page 274 or “The Amplitude Detector”
on page 283. After setting the parameters, and clicking the OK button, the
detections (if there are any) should appear in the sound window. Figure
10.2 shows the Amplitude Detector configuration dialog.
255
Figure 10.2 Amplitude Detector configuration dialog
Figure 10.2. The configuration dialog for the Amplitude Detector. After
setting the detection parameters, choosing OK will run the detector.
General Detector These parameters appear in the top half of the Configure Detector dialog
Parameters and are the same regardless of which detector you are running.
Presets
A preset includes a set of predefined parameter settings. Under the
Presets menu in the configuration dialog, you can apply an existing preset
or save your current settings for future use. Raven comes with a default
preset and a BlackCappedVireo preset for both the amplitude and energy
detectors. You can also create your own presets for various applications.
For more information on managing presets, see “The Preset Manager” in
Naming
Under the Name tab of the configuration dialog, you can choose a name
that will identify the detector in the side panel and become the name of the
256
created selection table. You can also opt to add a new annotation column
to the created selection table(s) with this name which will remain with the
selection if it is saved or moved to a different table. This can be useful if
you are running multiple versions of a detector and would like to better
differentiate the selections.
Tables
Under the Table tab, you can set the detector to automatically save the
resulting selections periodically. You can choose to automatically save the
selection table at the end of a detector run, or choose to automatically save
the table after every N detections. You can also select whether or not you
would like Raven to create a new selection table after each save. This will
result in several smaller selection tables, which may be easier to work with
than one large one. Finally, you can configure the naming and location of
the resulting files.
In the box labeled File Names, you can use a template to customize the
names your detector’s selection table files will be given. For more
information on how to use file naming templates, see “Sound File(s)” in
Chapter 4 (page 89). All of the tags listed in that section are available, and
you can also use the following tags:
Table 10.1. Additional template tags available in the General
Detector Parameters.
Tag
Definition
<f>
Filename of the sound file on which the
detector will run
<t>
Integer sequence number of the table.
Only relevant if you choose to create a
new table after every automatic save.
Channels and Timing
For a multi channel sound (or multi channel recorder) you can choose to
run the detector on all channels or a specific subset. You can also choose to
run the detector on the entire sound, or on a shorter portion of the entire
sound.
Displaying detector Once you’ve run the detector, controls for the detector are located on the
information Detection tab in the side panel. Also, if you make the selection table view
of the current sound visible, the name of the active detector will appear on
a new tab. The tab will be color-coordinated with the detector’s selection
boxes in the sound window and will display detections (selections created
by the detector) in the selection table. See Figure 10.3 for an example.
257
Figure 10.3 Detector tab
Figure 10.3. The detection tab, activated in the side panel, shows
information about the current detector. Also, detection information is
shown on its own tab in the sound window’s selection table.
You can add and remove measurements from the detection selection table
in the same way as you would with the default selection table. You can
also rename the tab by opening the tab’s context menu and selecting
258
Rename Table. See “Selection Tables” in Chapter 6 (page 150) for more
information.
After running a detector, the detector selections tab will
become the current active tab. This means that if you make a
manual selection in the sound window, the data will be
added to the detector tab and not the default selections tab.
To add selections to the default tab, you must make it the
active tab by clicking on it before making selections
Running an To test a detector on a long sound, first open the sound in a paged sound
interactive detector window, then configure an interactive detector as described previously.
in a paged sound This will run the detector on the current page. To see the results on
different pages in the sound recording, use the page scroll bar at the top of
the screen to advance one page in the recording. The detector will then
create selections for that page based on the input detection parameters.
Once you are satisfied with the results of a detector, you can choose to run
it on an entire recording. To do this, you can press the Run Full button in
the corresponding box on the detector tab in the side panel
.
Figure 10.4. The detector box as it appears for (a) a paged sound, and
(b) a non paged sound. For a paged sound, the detector active and
clear selection options and the Run Full button appear.
Detection options If you display the contents of the detector tab in the side panel, you will
see some boxes containing basic information about the detector(s)
259
currently open. These boxes also give you the ability to access or change
some other detector information.
Configure: This button allows you to change the detector parameters via
the standard configure dialog box. Once the parameters are changed and
the OK button is pressed, the detector will be run again on the currently
displayed page or the entire sound if it is all in memory. Note that
reconfiguring the detector will cause all current selections to be deleted.
Run Full: Pressing this button causes Raven to run the detector, with the
current set of parameters, on the entire sound recording. The results are
stored in the selection table associated with this detector. While the
detection is running, the Configure button will be grayed out. When the
detection is completed, the Configure button will become accessible again.
Detector Active: Checking this item box means that the detector will be
run on each page as it is opened. Unchecking the box means that the
detector will not run on subsequent pages as you display them.
Clear Selections: Interactive detectors are one way to test a set of
parameters. If testing parameters is your only goal, you may not want to
have all selections from every page saved. Checking the Clear Selections
box forces Raven to clear all old selections from the detector’s selection
table each time the detector runs. Note that reconfiguring the detector will
clear all selections, even if the Clear Selections box is unchecked.
The Run Full button, Detector Active box, and Clear Selections box are only available when a detector is configured in
a paged sound. Otherwise, you will not be able to select these
options
Running multiple You can configure multiple interactive detectors and run them
interactive simultaneously in Raven. Each will be given its own tab in the selection
detectors table with a corresponding color for the selections in the sound window
(Figure 10.5).
260
Figure 10.5 Multiple detectors
Figure 10.5. Two detectors that have been run on the same
spectrogram view. The first detector (green selection table tab and
green selections) is in a higher frequency range than the second
detector (pink tab and selections).
Full detection mode
Once you are satisfied with the results of a detector, you can choose to run
it on an entire recording. To do this, you can press the Run Full button in
the corresponding box on the detector tab in the side panel. Or, you can
use the Tools > Detector menu to start running a detector on an entire file
or file sequence.
261
An alternative is to bypass interactive mode and run the detector on the
full sound initially. Use the Tools > Detector menu to choose the detector
you want. The dialog in Figure 10.6 will appear, allowing you to choose an
existing file or open a new file (Raven will open any sound that is not
already open). This mode of detection can be used to post-process
recordings; by first recording to a file sequence, you can then run a
detector on the entire sequence at once.
Figure 10.6 detection input dialog
Figure 10.6. The detection input dialog for the Amplitude detector. This
allows you to choose the full sound on which you’d like to run the
detector. For other detectors the dialog may look different: it may ask
specifically for a waveform or spectrogram signal.
Batch detection
Once you come up with a set of detector parameters that are effective at
finding a certain type of signal, you may wish to run the detector on many
sound files at once to look for that kind of signal. Raven’s batch detector
allows you to automatically run the same detector on an arbitrarily large
number of files and save the results. Once you set up the batch detection
process, it runs in the background so that you can continue to do work in
Raven while it progresses.
262
To run the batch detector, choose Tools > Batch Detector... from the menu
bar. This will display the batch detection dialog (Figure 10.7).
Figure 10.7
Figure 10.7. The batch detection dialog.
Choosing files To perform a batch detection run, begin by choosing the files on which
you’d like to run the detector. Click the Add button to select the files you’d
like to add. You can select a single file by clicking on it, or multiple files by
holding down Control or Shift while you click. You can add more files by
clicking Add again. The files you choose do not have to be from the same
folder, and they do not need to be in the same format, as long as they are
all compatible with the detector parameters that you specify.
After you’ve added files, their names will show up in the Files box. If
you’d like to remove one of the files you added, click on its name in the
Files box and then click Remove. You can also remove all the files currently in the Files box by clicking Clear All.
When choosing files, you can add either individual sound files or list files.
For more information about list files, see “Using a list file” in Chapter 7
263
(page 195). If you add a list file, the detector will run back-to-back on all
the files in the list, and save a single selection table with the results from
all of those files.
Choosing a Once you’ve chosen your files, choose which type of detector you’d like to
detector run using the dropdown menu labeled Detector. You can choose between
the Amplitude Detector and the Band Limited Energy Detector. For more
information about those detectors, see “The Amplitude Detector” in
Chapter 10 (page 283) and “The Band Limited Energy Detector” in Chapter 10 (page 274).
Configuring the After you’ve chosen your files and your detector, you can configure the
detector detector by clicking the Configure Detector button. This will display the
detector configuration dialog. The dialog displayed will differ based on
whether you’ve selected the Amplitude Detector or the Band Limited
Energy Detector. If you’d like to use a detector configuration you’ve set
up before, or save the detector configuration you’re currently using for
future use, you can use the Preset menu (see “Presets” in Chapter 10 (page
256)).
For more information about how to configure the detector you’ve chosen,
please see “Configuring the Amplitude Detector” in Chapter 10 (page 283)
or “Configuring the Band Limited Energy Detector” in Chapter 10 (page
274).
Saving the To choose where to save the results of your batch detection run, click on
detection results the Table tab in the top panel of the detector configuration dialog (Figure
10.8). Note that the checkbox marked Automatically save table after detector run is already checked. During the batch detection run, the detector
will create a selection table of results for each file you select, and save each
of those tables in the place that you specify. Choose the folder in which
those results will be saved by clicking the Browse button next to the Save
Directory box. Then enter a template for the tables’ names in the File
Names box. We recommend that you use the symbol <f> somewhere in
your template; in the file names that are created, that <f> will be replaced
with each sound file’s file name. For more information about file name
templates, see“Tables” in Chapter 10 (page 257).
Warning: if you do not include the <f> symbol in your template, then multiple selection tables may be given the same
name. In that case they will write over each other, and some
of the results from your batch detection run will be lost.
264
Figure 10.8
Figure 10.8. The Table tab of the detector configuration dialog.
Configuring the If you choose the Band Limited Energy Detector, then the Configure View
view button will become available. Clicking it will display the spectrogram
configuration dialog. Configuring the view is optional; if you choose not
to configure the view, the batch detection run will use the default spectrogram configuration. If you do choose to configure the view, see “Configuring spectrographic views” in Chapter 5 (page 112) for more information.
Running the batch Once you’ve chosen a list of files and a detector, configured the detector,
detector and optionally configured the view, you’re ready to begin the batch detection run. Click theOK button at the bottom right of the batch detection
dialog. If the detector configuration you’ve chosen is not compatible with
the first file in your list of files, or if you’ve configured the view in a way
that makes your detector configuration invalid, a warning will pop up
asking you to change your configuration. Once these issues are resolved,
the batch detection run will begin.
When the batch detection run begins, Raven will display the Progress
Manager (Figure 10.9). The first panel in the Progress Manager, labeled
“Starting X detectors...”, keeps track of the batch detection run as a whole.
When the progress bar in that panel is complete, that means that individual detection runs have been started on each of the files you’ve chosen. If
some of the files have to be skipped, either because Raven is unable to
open them or because they are not compatible with the detector configuration you’ve chosen, their names will appear as a list of skipped files in the
first panel.
As each individual detection run begins, a new panel for that run will
appear in the progress manager. That panel will track how many detec-
265
tions have been made, how much data has been processed, and the detection speed so far. When the progress bar in one of those panels is
complete, that means the detector has finished running on the file listed at
the top of the panel.
Figure 10.9
Figure 10.9. The Progress Manager at the end of a batch detection
run. Note the list of skipped files in the first panel.
When all the progress bars have completed, the batch detection run is finished. The results of each detection are stored in the folder that you
selected in the Table tab of the detector configuration dialog.
At any point during the batch detection run, you can continue using
Raven for other tasks. If you’d like to make the Progress Manager visible
again, select Window > Progress Manager from the menu bar.
266
Real-time detection
Running a detector You can also run a detector within a recorder. To do this, first open a
within a recorder recorder window, and then configure an interactive detector that you
would like to run. Next, start the recorder and watch as the detector finds
signals according to the configured parameters, and creates corresponding
selections in the recording.
There are three ways to process detections from a recording:
1. Post-processing: This process involves running a recorder to a file
sequence, and then running a full detector on the file sequence after the
recorder has finished running. For this, you must record to a file
sequence completely, open the file sequence in a sound window, and
then run a detector on the file sequence.
2. Real-time processing while saving the entire recording: In this case,
you will save the entire recording but will run the detector while the
recorder is running. You can then save the resulting selections as a
selection table that can be opened in the saved sound, allowing you to
review the detections within the larger context of the sound at a later
time. To set up this type of recording session, run the recorder to record
to a file sequence, making sure to record to file (red button).
To ensure that the times in the selection table correspond to
the times in the file sequence, it is important to start recording
directly to a file sequence. Starting to record a file sequence
after recording to memory for a short time will result in a discrepancy between these times as the time shown in the
selection table corresponds to the time since the start of
recording, not the time since the start of the file sequence.
3. Real-time processing without saving the entire recording: Here, you
set the recorder to record to memory and you also configure the
recorder to save any selections as sound clips. With this method, you
are not saving the entire recording; you are only saving those selections
that the detector identifies. Unlike saving the detections as a selection
table within a saved file sequence, this method saves only the
detections, without the surrounding signal. Therefore, this method will
not allow any post-processing on non-selected data, so you should only
use this technique when you are confident that the detector is working
as anticipated to avoid losing data.
Example of running This example shows, step by step, how to run a recorder with a configured
a detector within a detector that saves clips to files, and then how to open and browse the
recorder results.
267
First, configure a recorder to record from your selected audio input
device. Configure it to record to memory and to save clips to files. The
exported clip files will be saved as individual sound files and as a list file
(see Figure 10.10.).
Figure 10.10
Figure 10.10. Configuring a recorder to record to memory and save clip
files to disk. Note that the names of the files are saved to a list file.
Next, configure a detector to find the signals that are of interest to you. In
this example we configure a detector to find short whistles. These settings
are shown in Figure 10.11. (For more information on these parameters, see
the next section, “The Band Limited Energy Detector” on page 274.). After
you configure the detector, a new selection table will be automatically
created in the recorder window and a box for the detector will appear in
the Detection tab of the side panel as shown in Figure 10.12.
268
Figure 10.11
Figure 10.11. The Target and Noise Tabs in the Configure Band
Limited Energy Detector dialog with parameters to find short whistles
269
.
Figure 10.12. A configured recorder with a selection table created for
the detector.
Turn on the recorder and begin recording a sound that includes some high
whistles. Notice that as the detector identifies these whistles, they are
marked in the spectrogram view, appear in the selection table, and are
also exported to clip files in real time
270
.
Figure 10.13. A running detector saving detected clips to files.
To view the clip files and confirm that they are saved correctly, open the
saved list file by going to File > Open Sound Files... and selecting the list
file. If the default naming conventions were used when configuring the
detector, this file is saved to the Clips folder within the Raven folder and
will be named with the date and time of recording (see Figure 10.14.).
When prompted, choose “Open as file sequence in one window” to view
all the detected selections in sequence. The clips should be opened in a
new sound window similar to the one in Figure 10.16.
271
Figure 10.14. Selecting a clip list file to open.
Figure 10.15. Choosing to open the list as a file sequence as opposed
to opening each sound in a separate window.
272
Figure 10.16
Figure 10.16. The clip files, opened in a file sequence.
Saving detected selections
Once you are satisfied with the selections created by your detector, you
can save your selections as a text file by selecting File > Save Selection
Table name]” As... or File > Save Selection Table “[Table name]”. In the
Save dialog, you must provide a name and path for the text file to be
saved. The default folder is the Selections folder within the Raven folder.
To save sound clips of the selections, you must select File > Save All
Selections As... or you can use the shortcut <Ctrl-Shift-A> (Windows,
Linux), or <Command-Shift-A> (Mac OS). Detections can also be saved
while the detector runs by configuring the recorder to export selections as
clip files and then running the detector. As the detections get marked as
selections, the recorder exports these selections as clips.
In addition to saving the detections, you may wish to save the detector
settings as well, in case you wish to run the detector again on the same or
different data. Some Raven detectors provide a preset menu to allow you
to save parameters. If a preset menu is not provided, then you should
record these settings on your own.
273
The Band Limited Energy Detector
Band Limited The Energy Detector supplied in Raven is a time-frequency energy
Energy Detector detector. It estimates the background noise of a signal and uses this to find
(Energy Detector) sections of signal that exceed a user-specified signal-to-noise ratio
threshold in a specific frequency band, during a specific time. First, the
detector makes a rolling estimate of the background noise, which is then
used to identify potential detections. Next, the detector compares
individual frames of a signal to the running noise estimate, selecting those
that exceed the specified threshold as candidates for detection. Finally,
Raven analyzes the distribution of these selected frames to identifiy
regions that satisfy the detector’s duration and occupancy parameters. The
energy detector runs on a spectrogram view only. In order to run it, the
associated view must be open or Raven will prompt you to open the
required view before running the detector.
Since this detector makes a rolling estimate of the background noise in the
band of interest, it is more useful when there is more uniform or “white”
noise, because the estimate can provide an adaptive threshold from which
real signal can be estimated. In more detail, the noise estimation process
begins by summing up the energy in the band for each frame in the
previous “block size” seconds. Then, it takes this list of energies and sorts
them before picking one according to the “percentile” parameter. For
example, a “30” in the “noise estimate percentile” parameter will take the
energy value 3/10th of the ways up from the bottom of the list. Likewise, a
value of “50” would result in the median value being chosen. This number
is then taken to be the noise estimate for that block, at which time the
“block” is advanced according to the hop size and the process is repeated.
So, if the noise block size is 5 seconds and the hop size is 1 second, the next
noise estimate will re-use the last 4 seconds, plus one more. This gives a
smooth “rolling average”.
A few consequences of this procedure are that a broad band setting will
result in a very small range of energy values for a block, resulting in a very
small difference between the noise estimate and the “highest” energies.
Also, to make a selection, the in-band energy needs to exceed the noise
estimate (by the SNR threshold amount) and then drop below it again
according to the duration/separation parameters. So, a noise estimate that
is too low can result in fewer selections, as can one that’s too high. This
makes the “percentile” number very important.
Configuring the To configure and run the energy detector, go to View > Interactive
Band Limited Detectors > Band Limited Energy Detector. This will open the Configure
Energy Detector Band Limited Energy Detector dialog box, which allows you to adjust the
detector parameters to isolate specific signals. In the upper portion of the
dialog box, you can set the general detector settings as described in
“General Detector Parameters” on page 256. The lower portion of the
dialog contains settings specific to the energy detector, which are
described below.
274
Target signal parameters
Under the Target tab, you can specify a range of frequencies on which to
run the detector and set length requirements for the detected signals.
Minimum and Maximum Frequencies (Hz, kHz): These determine the
range of frequencies in which the detector searches. The frequency band
should be wide enough to include any potential detections, but narrow
enough to exclude large bands of unnecessary noise. Since potential
detections are compared with the total noise level in the band, a wider
band setting can cause the detector to be less effective.
Minimum and Maximum Duration (sec, msec, samples): These
parameters specify the length of signal that can be considered a single
detection. These values should be set to include the full length of
detections you wish to identify. However, choosing a higher value for the
minimum duration reduces the likelihood that the detector will identify a
short segment of noise as a target signal. Analogously, choosing a lower
maximum duration prevents sustained tones in the frequency range from
being classified as detections.
Minimum Separation (sec, msec, samples): The minimum separation
corresponds to the amount of time between adjacent detections. If your
desired selections include short groups of sounds, this parameter can be
used to determine whether the individual elements are grouped together
as a single detection, or classified as multiple smaller ones.
Noise Estimation Parameters
The settings in the Noise tab control how Raven estimates the background
noise level. They also allow you to specify how loud a detection must be
with respect to the background noise.
Minimum Occupancy (%): The minimum occupancy defines the
percentage of samples within a selection that must exceed the background
noise by the SNR threshold in order for the signal to be considered a
detection. This prevents a brief section of louder noise from being
considered a detection when only a small portion of it exceeds the SNR
threshold.
Signal to Noise Ratio (SNR) Threshold (dB): The SNR threshold
establishes how far above the background noise a sample must be to be
considered a signal. Whereas the minimum occupancy determines how
much signal must be present in the detector’s range, the signal to noise
ratio defines the energy level that constitutes a signal above the
background noise.
Noise Block Size (sec, msec, samples): The block size corresponds to the
width of the block used to calculate the background noise level. Choosing
a block size substantially longer than the duration of a potential detection
prevents Raven from calculating noise from a selection that contains
mostly signal. However, if the background noise in your sound varies
275
considerably, a shorter block size will better account for this variation.
Note that if you attempt to run the detector on a segment of time smaller
than the noise block size, it will return no results.
Noise Hop Size (sec, msec, samples): The hop size determines how far the
noise block steps each time it recalculates. As a result, a shorter hop size
will result in more smoothing of the background noise levels. Choosing a
hop size longer than a typical detection, but shorter than the block size
generally provides the best results.
Noise Percentile (%): The noise percentile setting determines the
percentage of ranked noise that is counted as background noise. A higher
noise percentile setting results in a higher noise estimate, whereas a lower
noise percentile results in a lower noise estimate. The appropriate value
for a given application will depend on the specific characteristics of your
sound.
Exclusion band parameters
Raven can exclude selections that might otherwise be considered
detections through the use of an exclusion band. You can choose to ignore
selections that exceed the background noise by a given level within a
specified frequency range. This may be useful if your sound contains
signals that closely resemble the desired detections.
Minimum and Maximum Frequencies (Hz, kHz): These determine the
range of frequencies that are included in the exclusion band. You can
choose to exclude signals within a subset of the overall detection range, or
those with energy in a specific range outside the range of the detector. It is
possible to introduce new detections by using an exclusion band within
the range of the detector. This is because some selections that did not
occupy enough of the bandwidth before may now exceed the minimum
occupancy within the reduced frequency range.
SNR Threshold (dB): The SNR threshold in the exclusion band affects
how sensitive the detector is to signals in the exclusion band. A low
threshold value excludes potential detections which contain even a weak
signal in this range, whereas a higher threshold value only excludes
selections with considerable energy in the exclusion band.
Bandwidth filter parameters
The bandwidth filter in the energy detector allows you to target selections
with particular frequency distributions. Using these settings, you can
choose how wide of a frequency band individual selections should
occupy. This feature can be useful when searching for detections that
occupy a narrow bandwidth, but may be located across a larger frequency
range.
Minimum and Maximum Bandwidth (Hz, kHz): These values
correspond to the allowed bandwidths for individual detections. Setting
276
these parameters determines the minimum and maximum range that can
be occupied by the amount of energy specified in the energy percentile
setting. For example, specifying an energy percentile of 50% and
minimum and maximum bandwidths of 100 Hz and 200 Hz respectively
would require that 50% of the energy in each detection be concentrated in
a range wider than 100 Hz, but narrower than 200 Hz. However, this span
may occur anywhere in the detector’s range
Energy Percentile (%): The energy percentile corresponds to the
percentage of the total energy in a signal that must be concentrated within
the specified range of bandwidths.
Band Limited To demonstrate the use of the Band Limited Energy Detector, we will walk
Energy Detector through an example, demonstrating how to configure an effective set of
Example detector parameters for the file BlackCappedVireo.aif. Raven comes with
these parameters saved as the BlackCappedVireo.energy preset. Figure
10.17 shows the Band Limited Energy Detector configuration dialog with
this preset loaded.
277
Figure 10.17
Figure 10.17. The Band Limited Energy Detector configuration dialog
with the BlackCappedVireo.energy preset loaded.
Choosing the minimum and maximum frequencies is straightforward:
choose an example of a typical event you wish to detect. Make a selection
which encompasses a significant fraction of the energy of this event, and
read the Low Freq and High Freq measurements from the selection table.
In the case of the Black-capped Vireo sound, the goal is to detect several
different types of events in one detector run. To achieve this, draw
278
selections around each of the event types you wish to detect, and set the
minimum frequency equal to the lowest Low Freq measurement obtained,
and the maximum frequency equal to the highest High Freq measurement
obtained. See Figure 10.18.
Figure 10.18
Figure 10.18. Finding the minimum and maximum frequencies. Note
that the least Low Freq measurement is around 1500 Hz and the
greatest High Freq measurement is around 10000 Hz. These
correspond to the preset’s minimum and maximum frequency values.
To choose values for the minimum and maximum duration parameters,
add the Delta Time measurement to your selection table. Find the shortest
duration event you wish to detect and draw a selection around it. Read the
Delta Time measurement from the selection table, and use this value for
the minimum duration (for an example, see Figure 10.19). To obtain the
279
maximum duration parameter, repeat the procedure with the longest
duration event you wish to detect.
For settings with a time dimension, Raven automatically
rounds values entered in seconds or milliseconds to the nearest frame size. For example, a value of 1 second may automatically be rounded to 1.00136 seconds. This change will
appear immediately after the affected setting is entered.,
Figure 10.19
Figure 10.19. Finding the minimum duration. Note the minimum Delta
Time measurement of 0.101 seconds.
To choose a value for the minimum separation parameter, find the two
closest distinct events and draw a selection between them. Read the Delta
Time measurement from the selection table, and use this value for the
minimum separation.
280
Figure 10.20
Figure 10.20. Finding the minimum separation.
Skip over the minimum occupancy and SNR threshold parameters for
now, and set the Noise Power Estimation parameters. Pick a block size
that is longer than the maximum duration by at least a factor of 3, and
longer if the events cluster together. The goal is to have a long enough
block size that the data being used for noise estimation is guaranteed to
always include noise. If you make the noise block too short, it may result
in noise estimates based primarily on only signal, rather than signal and
noise together. In the BlackCappedVireo.energy preset, the block size is
about three times the maximum duration.
The hop size parameter determines how much smoothing is performed in
the noise estimation computation. Choose a hop size value somewhere
between the maximum duration and the noise block size. The
BlackCappedVireo.energy preset has the hop size set to 0.499 seconds,
slightly less than the maximum duration of 0.597 seconds.
An appropriate value for the noise percentile parameter depends on the
characteristics of the noise and clutter in your data. A higher percentile
will result in a higher noise estimate. For the BlackCappedVireo.energy
preset, the default 20th percentile was used, and appears to give a
reasonable noise estimate. You may wish to play with higher and lower
values to see which gives the best result for your data.
Once the Noise Power Estimation parameters have been determined, go
back to the SNR threshold parameter. For an approximation of this
281
parameter, you can measure the average power of a typical noise block.
Figure 10.21 shows a typical detection and a typical noise block. The
difference in average power is 10.5 dB. The BlackCappedVireo.energy
preset rounds this to 10 dB.
Note that the noise block in Figure 10.21 actually contains a large amount
of signal. This is why the SNR threshold as computed here is so low. Since
Raven uses all the sound in each block to estimate background noise, both
the distribution of signal within the sound and the block size will affect
the calculated SNR. In this example, the calculated SNR is significantly
lower than the true SNR due to the rapid succession of calls.
Figure 10.21
Figure 10.21. Measuring the estimated signal-to-noise ratio. Selection
1 contains only signal. Selection 2 represents a typical block of data
used to estimate the noise power.
Accept the default minimum occupancy, and check the Return Occupancy
checkbox. Then run the detector. Examine the detections and determine if
the performance was satisfactory; if not, adjust the minimum occupancy
downward and run the detector again. Sorting the selection table by
occupancy can serve as a guide to setting the minimum occupancy
parameter. Note that the BlackCappedVireo.energy preset has a minimum
occupancy of 40%, yet it yields detections with occupancies no lower than
72.97%. The preset’s minimum occupancy could have been set as high as
72% and returned the same results.
282
When citing use of Raven’s Band Limited Energy detector in scientific
publications, please refer to it as follows:
Mills, H. G. 2000. Geographically distributed acoustical monitoring of
migrating birds. Journal of the Acoustical Society of America, 108(5, part 2):
2582.
The Amplitude Detector
The Amplitude Detector detects regions of a signal where the magnitude
of the waveform's envelope exceeds a threshold value. To approximate the
envelope, the detector computes the absolute value of the waveform and
passes it through a lowpass smoothing filter. The envelope signal is then
tested against the threshold. Raven does not display the actual envelope
data.
Configuring the To configure the detector, first choose the amplitude threshold based on
Amplitude Detector the waveform amplitude. Then set the smoothing time constant equal to
the most rapid rise or fall of the waveform that you would like the
envelope detector to follow (see Figure 10.22). After running the detector,
you may notice detections that occur slightly after the waveform actually
crosses the desired threshold. To minimize this delay, set the delay
compensation parameter to the size of the offset.
283
Figure 10.22 configure detector dialog
Figure 10.22. The dialog box to configure an amplitude detector.
To demonstrate the use of the amplitude detector, we will walk through
an example. First, open the ChestnutSidedWarbler.aif file (included in the
Raven Examples folder). Next, we will choose the amplitude threshold by
moving the vertical position marker until it intersects a set of events
typical of those which you would like to detect, but above the noise and
clutter you would like to reject. Then, read off the amplitude threshold
value from the y-axis on the left. In this example, the reading is 5.015
kUnits (or 5015 Units). See Figure 10.23.
284
Figure 10.23 the amplitude threshold
Figure 10.23. The ChestnutSidedWarbler.aif sound window with the
vertical position marker (in the waveform view) positioned at 5015
Units. This reading will become the amplitude threshold for our
detector.
Next, make a selection which encloses the most rapid rise or fall of the
waveform that you would like the envelope detector to follow. Ideally, the
rise should be from zero to the maximum value of the waveform. In this
example, we’ve used the Delta Time measurement to compute the width
of the selection for us. The time width, τ, is .09575 s which we will use as
the value of the smoothing time constant. See Figure 10.24.
285
Figure 10.24 calculating the time width
Figure 10.24. After selecting the most rapid rise/fall of the waveform
(selection 1, in red) we use the Delta Time measurement (highlighted in
purple in the selection table) to obtain a value for the smoothing time
constant.
After running the amplitude detector on ChestnutSidedWarbler.aif with
the previously determined settings (amplitude threshold of 5015 Units,
smoothing .09575) you should get something similar to Figure 10.25. The
detections generally correspond to the regions we wish to detect, but are
slightly delayed as illustrated in Figure 10.26. The delay compensation
parameter minimizes this delay by shifting the detections earlier by a
specified amount. To use this feature, make a selection spanning the time
between when the waveform crosses the amplitude threshold and the
start of a detection and use the delta time measurement as the delay
compensation parameter.
286
Figure 10.25 detections
Figure 10.25. The selections made by the amplitude detector.
Figure 10.26Delay compensation
Figure 10.26. A magnified view of the selections made by the
amplitude detector. Notice the delay between the time when the
waveform crosses the threshold and the start of the detection.
287
Once the delay compensation parameter has been set to the appropriate
value, the resulting selections should better align with the actual
waveform, as shown in Figure 10.27.
Figure 10.27 Amplitude detector with delay compensation parameter
Figure 10.27. Selections made by the amplitude detector using delay
compensation parameter of 0.009 seconds. The detections now
correspond more accurately to the shape of the waveform.
288
Chapter 11
Customizing Raven
About this chapter
This chapter explains how to use Raven preferences to configure many
aspects of Raven’s default appearance and behavior. We discuss general
Raven preferences and specific information on the following topics:
•editing Raven color schemes
•the preset manager
•using the Raven preferences file
•power spectra calculation methods
•default Raven settings and presets
•the memory manager
•Raven updates
•troubleshooting and performance tips
•contacting the Raven development and support team
Editing color schemes
It is possible to customize the color scheme used in Raven. Figure 11.1
shows an example of two color schemes that differ in spectrogram color
map, background color, axis color, and waveform line color.
289
Chapter 11: Customizing Raven
Figure 11.1. Varied color schemes
Figure 11.1. The same sound window displayed with two different color
schemes.
290
By using the Color Scheme Editor, you can change any of the color scheme
elements listed below.
• 1st Quartile Frequency
• line title
• 1st Quartile Time
• max amplitude
• 3rd Quartile Frequency
• max bearing
• 3rd Quartile Time
• max frequency
• active point selection
• max power
• active selection border
• max time
• active selection fill
• min amplitude
• active selection label
• min time
• average power
• no data
• axis
• peak amplitude
• background
• peak correlation
• center frequency
• peak frequency
• center time
• peak lag
• Energy
• peak power
• file separators
• peak time
• filtered RMS amplitude
• playback cursor
• frequency 5%
• position label
• frequency 95%
• position marker
• gram slice line
• RMS amplitude
• grid (major)
• selection spectrogram line
• grid (minor)
• time 5%
• inactive point selection
• time 95%
• inactive selection border
• time series line
• inactive selection fill
• waveform dot
• inactive selection label
• waveform line
291
The inactive selection colors in the color scheme set the colors for the default selection table. For other tables, the inactive point selection, inactive selection border, and inactive
selection fill are set to the individual table color. The inactive
selection label color and inactive selection fill opacity are set
in agreement with the default table as prescribed in the color
scheme. Also see “Opacity” on page 294.
To open the Color Scheme Editor (Figure 11.2), choose Color Scheme >
Edit... from the View menu or from the contextual menu for any view.
Figure 11.2. Color scheme editor
Figure 11.2. The Color Scheme Editor dialog. The Swatches color
chooser is displayed by default when the Color Scheme Editor is first
opened.
292
Choosing a color To change the color of a particular element of a color scheme, choose the
scheme element to name of that element from the Color drop-down menu, and then set its
edit color using one of the three color choosers that appear on the tabs in the
middle of the dialog box.
Choosing colors Raven’s color choosers provide three different ways of specifying colors,
as described below. For all color choosers, a sample of the color chosen is
shown in the Preview panel in the lower third of the dialog box.
Swatches color chooser
When you first open the Color Scheme Editor dialog box, the Swatches
tab is displayed by default (Figure 11.2). To change the color of the
selected color scheme element, click on one of the small color swatches in
the rectangular color palette. Each time you click on a swatch, a swatch of
that color is added to the small Recent palette to the right of the main
palette. You can choose any color that you have used recently either from
the Recent palette or from the main color palette.
HSB color chooser
The HSB color chooser (Figure 11.3) allows you to choose colors for each
element in a color scheme by specifying the element’s hue, saturation, and
brightness. You can set the hue, saturation, and brightness either by typing
numerical values into the H, S, and B text fields, or by using the graphical
controls.
Figure 11.3. Color scheme editor-HSB
Figure 11.3. The HSB (hue, saturation, brightness) color chooser in the
Color Scheme Editor dialog.
Hue: Values for hue (H) must be between 0 and 359. The current hue is
indicated by the position of the slider control to the left of the vertical color
bar. To change the hue, move the slider up or down, or type a new value in
the H field and press <Enter>.
293
Saturation and Brightness: Values for saturation (S) and brightness (B)
must be between 0 and 100. The large square panel to the left of the hue
slider represents possible combinations of saturation and brightness. The
horizontal axis of the square represents color saturation ranging from 100
at the left edge to 0 at the right edge. The vertical axis of the square
represents brightness ranging from 100 at the top edge to 0 at the bottom
edge. To set the saturation and brightness of the selected color scheme
element, click on the point within the square that corresponds to the
desired saturation and brightness. You can also drag the small white circle
that indicates the current saturation and brightness within the saturationbrightness square, or type new values in the S and B fields, and press
<Enter> for each one.
RGB color chooser
The RGB color chooser (Figure 11.4) allows you to choose colors for each
element in a color scheme by specifying the relative amount of red, green,
and blue that are added together to yield the element’s color. Values for
red, green, and blue range between 0 and 255. You can specify RGB values
by typing them into their respective fields in the dialog, or by sliding the
corresponding controls left or right.
Figure 11.4. Color scheme editor-RGB
Figure 11.4. The RGB (red, green, blue) color chooser in the Color
Scheme Editor dialog box.
Opacity There is an ordering relationship among the various elements that are
displayed in a view such that any one element is either in front of or
behind any other element. Each element in a view has an opacity
associated with it that can vary between 0 and 255. An element that has an
opacity of 255 is completely opaque— it completely hides whatever
elements are behind it. An element with an opacity of 0 is completely
transparent, hence invisible. An element with an intermediate opacity
allows elements behind it to be seen, but tinted with the front element’s
294
color (Figure 11.5). By default, Active Selection Fill and Inactive Selection
Fill have opacity values of 15; all other elements are completely opaque
(opacity = 255). You can set the opacity of the selected view element by
moving the horizontal Opacity slider.
Selection tables other than the default (first) table will get their
color from the selection table color and will inherit their inactive selection fill opacity from the inactive selection fill color
opacity in the color scheme.
Figure 11.5. Opacity comparison
Figure 11.5. Two spectrogram views showing the effect of varying
opacity of the Active Selection Fill color. In the upper view, the opacity is
set to 15 (the default value); in the lower view it is set to 80.
Color scheme You can save the current color scheme as a color scheme preset, which you
presets can then apply later to any sound window, without having to specify the
colors of all of the elements in the color scheme individually. Commands
for saving and applying color scheme presets are in the color scheme
Preset menu, which is accessible from the View menu, from the contextual
menu for any view, and from within the Color Scheme Editor dialog box.
To save the current color scheme as a preset, choose Preset > Save As... .
To apply an existing color scheme preset, choose the name of that preset
295
from the color scheme Preset menu. When the Save Color Scheme Preset
dialog box appears, enter a name for the preset, and click OK.
Color scheme presets must be saved in the folder Presets/
Color Scheme/ within the Raven program folder. You can
also create additional folders within the Color Scheme
folder by clicking on the New Folder icon within the Save
Color Scheme Preset dialog. These folders will appear as
submenus in the color scheme Preset menu, with each submenu listing the presets in the corresponding folder.
Auto-Apply, Apply, If the Auto-apply checkbox is checked (the default), Raven immediately
Reset, OK, and applies any change you make to an element’s color or opacity, without you
Close needing to click the Apply button. For parameters that you enter by typing
in a field (e.g., Red, Green, or Blue values in the RGB color chooser), the
change is applied when you complete an entry by pressing the <Enter> or
<Tab> key, or by clicking in another field.
If the Auto-apply checkbox is unchecked, changes that you make in the
Color Scheme Editor are not applied until you click the Apply or OK
button. Clicking the OK button applies the changes and closes the dialog.
Clicking the Apply button applies the changes but leaves the dialog open.
Clicking the Close button closes the dialog without applying changes.
If you click the Reset button before applying changes, the color and
opacity settings revert to their previous values.
If Auto-apply is checked, the Reset button will not restore the
color and opacity settings in effect before you made changes.
Creating In addition to the predefined spectrogram colormaps, you can create
customized custom colormaps to use in Raven. A colormap consists of a text file that
colormaps specifies colors at any number of specified control points. A control point
corresponds to a particular decibel level at which the color is explicitly
defined. For points other than the control points, the colors shade evenly
from the previous control point's color to the next one.
The first line of the test file must contain a one word name for the
colormap and the total number of control points in the file separated by a
space. The following lines must contain the location of a control point
followed by the desired RGB color values, each separated by a space. Both
the control point locations and color values correspond to a percentage of
the available range and must be between 0 and 1. You can specify the color
at any number of control points by adding the corresponding number of
lines. When complete, the format of the file should be as follows:
296
[One-word map name] [number of control points]
[Control point 1's location] [R] [G] [B]
[Control point 2's location] [R] [G] [B]
etc...
For example a colormap called “Summer” that specifies colors at five
evenly spaced control points is provided below. Although the values have
been arranged in columns here for easier reading, in the actual file, they
must be separated by only a space.
Summer 5
0
0.976
0.25 0.533
0.5 0.243
0.75 0.572
1
0.933
0.953
0.623
0.345
0.372
0.767
0.922
0.831
0.67
0
0.008
After you have created a text file for your colormap, it must be saved in
the colormaps folder in the Raven program directory. Depending on your
installation of Raven, this folder may or may not already exist. If it does
not exist, you will need to create a new folder named “colormaps” within
the Raven program directory.
Once you have saved your colormap in the colormaps folder, it will be
available in Raven with the other colormaps under View > Color Scheme
or by right clicking in the sound window and choosing Color Scheme
from the contextual menu.
The Preset Manager
Table 11.1 lists all of the types of presets available in Raven. You create
presets by choosing Preset > Save As... in any dialog box that supports
presets.
297
Table 11.1. Types of presets available in Raven.
Audio File Format
Audio File Names
Audio Recorder
Audio Recorder Display
Audio Recorder Input
Audio Recorder Schedule
Clip File Format
Clip File Names
Color Scheme
Detector
Filter
Measurement List
Sample Rate Conversion
Selection Labels
Sound Window
Spectrogram Parameters
Speed Correction
You can use the Preset Manager (Figure 11.6) to delete or rename presets,
to create, rename or delete folders within preset folders, and to move
presets from one folder to another. To open the Preset Manager window,
do one of the following:
• choose Window > Preset Manager from the Raven menu bar
• choose Window Preset > Manage Presets from the View menu or the
contextual menu for any view
• choose Preset > Manage Presets from the menu bar in any dialog box
that supports presets.
298
Figure 11.6 Preset manager
Figure 11.6. The Preset Manager dialog box, showing the Spectrogram
Parameters preset selection.
If you access the Preset Manager from a Preset menu in a dialog box, it
will initially display presets of the appropriate type for the dialog box
from which you invoked it. To choose a different preset type to manage,
use the Preset Type drop-down list. The list box will show the names of all
of the presets and folders presently defined for the selected preset type. To
rename or delete a preset, click on its name in the list, then click the
appropriate button. To create a folder within the currently selected preset
type, click New Folder... . To move a preset or folder into or out of a folder,
simply drag its entry within the list.
About Raven preferences
You can control the default appearance and behavior of many parts of
Raven by setting preferences. You set preferences by editing the Raven
preference file with a text editing program or word processor.
Raven allows you to specify preferences for the following:
• Default presets for spectrogram parameters, recorder parameters,
window layout, view color scheme, and measurement list.
• Power spectra calculation method that Raven will use.
• Default workspace to open when Raven is launched.
• Default directories that Raven will use for opening and saving files.
• Default precision for measurements displayed in the selection table.
• Behavior of audio recorders when creating files. If a file created by an
audio recorder has the same name as an existing file, the new file can
299
either overwrite the old one, or be placed in a new directory in order to
preserve the old file, depending on the preference you specify.
• Return email address to use when sending feedback to the Raven
development team.
The Raven Raven preferences are specified in a text file named
preference file RavenPreferences.txt, which is placed in the Raven program
directory when Raven is installed. Each time Raven is launched, it loads
the preference file. To change preference settings, you must edit this file. If
Raven is running when you edit the preference file, you will need to quit
and restart Raven for your changes to take effect.
Figure 11.7.. The Edit toolbar with preferences icon indicated.
Before editing the preference file, it’s a good idea to make a
backup copy of the file, in case you inadvertently damage or
delete the file you’re editing.
To edit the file, double-click on its icon, and the file should open in your
computer’s default text editing program. Alternatively, you can open the
preference file from within any text editing program that can work with
plain text files (sometimes called text only or ASCII files).
When you edit the preference file, make sure that the revised
file is still named RavenPreferences.txt and that it is
saved as a plain text file. If your word processing program
asks you if you want to save the file in the program’s own
native file format instead, you should always choose plain
text (or “text only”).
Each particular preference is specified by an entry in the preference file of
the general form
raven.prefName=prefValue
where prefName is replaced by the name of the particular preference and
prefValue is replaced by a valid value for that particular preference. For
example, the entry
raven.preset.colorScheme.defaultPreset=Default
300
specifies that Raven should by default apply the color scheme preset
named Default to new views.
The character ‘#’ is considered the comment character in the preference
file— any text that follows this character up until the end of the line is
disregarded by Raven. You can thus add comments to a preference file to
make it more intelligible to a human, by preceding each comment line
with ‘#’. The preference file that is supplied as part of the standard Raven
installation contains comments that explain the meaning of each
preference line. You can also “comment out” an entry and cause Raven to
ignore it by inserting ‘#’ at the beginning of the line.
Although some preference entries are shown in this manual
across two lines of text, each entry in the actual preference
file must be on a single line.
The sections below describe each category of preference in detail.
Default presets
Presets provide a mechanism for saving sets of related configuration
parameters for Raven sound windows and views. Presets simplify
configuring windows and views by enabling you to set multiple
parameters in one step. For example, a color scheme contains color
specifications for the spectrogram color map and for fifteen other display
elements, such as point selections, border and fill for range selections,
position markers, and so on. You can alter any of the predefined colors
that Raven provides, and then save the entire set of colors as a color
scheme preset that can be retrieved later with a single mouse click.
The sections below briefly describe each type of preset, provide crossreferences to further discussions elsewhere in this manual, and explain
how to specify a preference for the default preset.
Spectrogram A spectrogram preset saves the state of all of the parameters and controls
presets in the Configure Spectrogram dialog under a single preset name. You can
retrieve a saved spectrogram preset from the Preset menu within the
Configure Spectrogram dialog. Spectrogram presets are discussed further
in “Spectrogram presets” in Chapter 5 (page 129).
Raven is installed with a spectrogram preset named Default, which is
initially specified as the default by the following entry in the preference
file:
raven.preset.spectrogramParameters.defaultPreset
=Default
301
To change the default to a different spectrogram preset, edit this entry. For
example, to use a spectrogram preset called MySpectrogram by default,
change the default spectrogram preset entry to
raven.preset.spectrogramParameters.defaultPreset
=MySpectrogram
Spectrogram views that are created by applying a window
preset will use the spectrogram parameters specified by the
window preset, not those in the default spectrogram preset.
Recorder presets A recorder preset saves the state of all parameters and controls on one or
all of the tabs in the Configure Recorder dialog under a single preset
name. You can retrieve a saved recorder preset from the Preset menu
within the Configure Recorder dialog. Recorder presets are discussed
further in “Recording to a file sequence” in Chapter 4 (page 92).
When Raven is first installed, there is no default recorder preset defined.
This is because a recorder preset includes the name of the default audio
input device, which varies among computers, and thus cannot be included
in a predefined recorder preset. To specify a default recorder preset, you
must first configure a recorder with the parameters that you will use most
commonly, and save that configuration as a recorder preset named
Default, as described in “Recording to a file sequence” in Chapter 4
(page 92). Then uncomment the following entry in the preference file by
deleting the ‘#’:
# raven.preset.audioRecorder.defaultPreset=Default
Measurement A measurement preset saves a list of measurements to be displayed in a
presets selection table, along with their respective formats and precisions. You can
retrieve a saved measurement preset from the Preset menu within the
Measurement Chooser dialog. Measurement presets are discussed further
in “Using measurement presets” in Chapter 6 (page 177).
Raven is installed with a measurement preset named Default, which is
file:
raven.preset.measurementList.defaultPreset=Default
To change the default to a different measurement preset, edit this entry.
For example, to use a measurement preset called MyMeasurements by
default, change the default measurement preset entry to
302
raven.preset.measurementList.defaultPreset
=MyMeasurements
When a window preset is applied to a sound window, the
measurements specified in the window preset take precedence over the default measurement preset. See “Sound
Window Presets” in Chapter 3 (page 71).
Sound window A sound window preset saves information about the layout of a sound
presets window, such as window size, what views are displayed, their sizes, color
schemes, position and scale, selection table visibility and measurements
displayed. You can retrieve a saved window preset from the Window
Preset menu on the View menu or the contextual menu for any view.
Sound window presets are discussed further in “Sound Window Presets”
Raven is installed with a sound window preset named Default, which is
file:
raven.preset.soundWindow.defaultPreset=Default
This preference actually uses the Default1.3 window preset
when the Raven 1.3 power spectra calculation method is
used. This change results in a default spectrogram that looks
the same despite the change in calculation method.
To change the default to a different sound window preset, edit this entry.
For example, to use a sound window preset called MyWindow by default,
change the default measurement preset entry to
raven.preset.soundWindow.defaultPreset=MyWindow
Color scheme A color scheme preset saves the colors of all elements in a color scheme
presets (the spectrogram color map, and colors of view elements such as position
markers, selection borders and fills, and background) under a single
preset name. You can retrieve a saved color scheme preset from the Color
Scheme > Preset menu on the View menu or the contextual menu for any
view.
Raven is installed with a color scheme preset named Default, which is
file:
raven.preset.colorScheme.defaultPreset=Default
303
To change the default to a different color scheme preset, edit this entry.
For example, to use a sound window preset called MyColors by default,
change the default measurement preset entry to
raven.preset.colorScheme.defaultPreset=MyColors
When a window preset is applied to a sound window, the
color scheme(s) specified in the window preset take precedence over the default color scheme preset. See “Sound
Window Presets” in Chapter 3 (page 71).
Selectable look and
feel, desktop
background color,
tooltip color
You can choose between several predesigned “look and feel” schemes that
change the appearance of the Raven program. On Windows and Linux the
list includes Metal, CDE/Motif, Windows, or Windows Classic. On Mac
OS, the list includes Metal, CDE/Motif, or Mac OS X (sometimes known as
Aqua). Raven 1.4’s standard appearance is Metal on Windows and Linux
and Mac OS X on Mac OS, but you can adjust the current scheme at any
time by choosing Window > Look and Feel.
You can also change the background color of the desktop by choosing
Window > Background Color... which will open the Background Color
Editor panel. You can set the color of the main Raven screen this way.
Raven will save the color in the preferences file so that it can be applied on
subsequent invocations of Raven.
You can also change the color of Raven’s tooltips, the helpful signs that
pop up when you hover the mouse cursor over a button or box. To choose
a tooltip color, choose Window > Tooltip Color... which will open the
Tooltip Color Editor panel. Note that tooltip color can only be changed
within the Metal Look and Feel.
Power Spectra Calculation Method
In Raven 1.3, a new power spectra calculation method was introduced.
The newer method includes energy in the negative frequency bins of the
DFT, as well as normalization of the power spectrum by the number of
points in the DFT window. This newer method should be used in cases
when measurements need to be consistent with those made in Canary.
To change the calculation method, edit the last line of text which should
read as either:
raven.compatibility.computation.spectrogram=1.2
or
304
This statement indicates the power spectra computation method that
Raven is currently using. To use the new method, change the entry to 1.3,
otherwise, use 1.2.
You can also compare the two calculation methods by simultaneously
opening two instances of Raven, one with the Raven 1.3 method active
and the other with the Raven 1.2 method active. To do this, you must:
1. Start Raven 1.4. This instance of Raven 1.4 will continue to use the
current method, which in most cases will be the one that was chosen
the first time that Raven 1.4 was started.
2. Without closing the first instance of Raven 1.4, start Raven 1.4 again so
that there are now two instances running simultaneously. This instance
of Raven is the one for which you will alter the calculation method.
3. With the second instance of Raven 1.4 active, open the Raven
preferences file by clicking on the preferences icon in the toolbar.
4. Scroll to the last line of the text which should read as either:
or raven.compatibility.computation.spectrogram=1.3
5. If the version is 1.3, change it to 1.2, or vice versa. This change will
affect only the power spectra computation method.
6. Save and close RavenPreferences.txt
7. Leaving the first instance of Raven 1.4 open, close, then reopen the
second instance of Raven 1.4.
8. The previous preference settings will continue to apply to the first
instance of Raven 1.4 as long as it remains open. So, it uses the power
spectra computation method that was initially set, while the second
instance uses the method that was just set.
Using the current Raven 1.3 power spectra calculation in one instance, and
the earlier Raven 1.2.x power spectra calculation in the other instance,
open the same file in each and view them simultaneously to see the
difference.
Default workspace
Workspace files are discussed in “Saving Your Workspace” in Chapter 1
(page 15).
You can specify a default workspace for Raven to open each time the
program is launched by uncommenting and editing the following entry in
the preference file
# raven.workspace.defaultWorkspace=./Workspaces/
Default.wsp
The notation “./” means that the remainder of the specified pathname is
within the Raven program directory. If you want to use a default
workspace that is located elsewhere, you must specify a complete path, as
in
305
raven.workspace.defaultWorkspace=C:/MyRavenStuff/
MyDefault.wsp
A default workspace is useful for changing the initial appearance of the
Raven window. For example, the main Raven window is by default
initially 750 pixels wide and 550 pixels tall. If you want Raven to always
start with a window that fills the entire screen, you can launch Raven,
expand the window to full-screen, and then save the workspace. Then edit
the preference file to specify that workspace as the default. Thereafter,
Raven will start with a full-screen window. In addition to window size,
the workspace will save information on the state of the side panel
(visibility, size, which tab is selected, and so on).
Remember that a workspace file saves the entire state of
Raven, including all of the sound windows that are open. If
you just want to save the size and layout of the main Raven
window, save the workspaces when no sound windows are
open.
Default directories
You can specify the default directories that Raven uses when you open
and save sound files, selection files, or workspace files. When Raven is
first installed, these defaults are set respectively to directories named
Examples, Selections, and Workspaces within the Raven program
directory. These defaults are specified by the following three entries in the
preference file:
raven.ui.openFileChooser.defaultDirectory=
./Examples
raven.ui.selectionFileChooser.defaultDirectory=
./Selections
raven.ui.workspaceFileChooser.defaultDirectory=
./Workspaces
The notation ‘./’ means that the specified directories are subdirectories of
the Raven program directory. You can change any of these to any
directory on your hard disk by editing the corresponding entry in the
preference file. For example, the following entry tells Raven to use the
directory c:/MyStuff/SoundFiles as the default directory for sound
files:
306
raven.ui.openFileChooser.defaultDirectory=
c:/MyStuff/SoundFiles
You need to use the forward slash (/) character to separate
directories and subdirectories in Raven preference entries,
not the back slash (\). Note that this is different than entries
in list files, for example.
Default measurement precision
By editing the appropriate entries in the preference file, you can specify
how many digits of precision Raven should use by default for displaying
and saving measurements in the selection table. You can always change
the precision of a measurement displayed in a selection table, as described
in “Measurement precision and format” in Chapter 6 (page 177).
If you invoke a measurement preset, either by choosing a
preset from the Preset menu within Raven’s Measurement
Chooser, or by specifying a default measurement preset in
the preference file, the measurement precisions saved for
measurements included in the preset will override precision
preferences. Measurement precision preferences will apply
only to measurements that are not specifically included in the
the preset, when and if these are added to the selection table
via the Measurement Chooser.
Measurement precision entries are of the form
raven.measurement.basic.name.defaultValueFormat
=precision
where name is replaced by the name of a measurement type or of a
specific measurement, and precision is replaced by a string such as
0.000. The number of zeros to the right of the decimal point indicates
the desired precision.
Raven provides measurements of four types of quantities: time, frequency,
amplitude, and power. For each type of measurement, you can specify a
default display precision. You can also specify a different precision for
specific measurements within each type, which will override the default
for the type. For example, if you were to specify a default precision of
0.000 for time measurements, and a specific precision of 0.00 for the
Delta Time measurement, Raven would display two digits to the right
of the decimal point for Delta Time values, and three digits for all other
time values.
307
Time To specify the default measurement precision for time measurements,
uncomment and edit the following entry:
# raven.measurement.basic.time.defaultValueFormat
=0.000
To override the default measurement precision for time measurements
and specify a different precision for Begin Time, End Time, Delta
Time, Max Time, Min Time, or Peak Time, uncomment and edit the
appropriate one of the following entries:
#
raven.measurement.basic.beginTime.defaultValueFor
mat=0.000
#
raven.measurement.basic.endTime.defaultValueForma
t=0.000
#
raven.measurement.basic.deltaTime.defaultValueFor
mat=0.000
#
raven.measurement.basic.maxTime.defaultValueForma
t=0.000
#
raven.measurement.basic.minTime.defaultValueForma
t=0.000
#
raven.measurement.basic.peakTime.defaultValueForm
at=0.000
Frequency To specify the default measurement display precision for frequency
measurements, uncomment and edit the following entry:
#
raven.measurement.basic.frequency.defaultValueFor
mat=0.0000
To override the default measurement precision for frequency
measurements and specify a different precision for Low Frequency,
High Frequency, Delta Frequency, or Max Frequency,
uncomment and edit the appropriate one of the following entries:
# raven.measurement.basic.lowFrequency.defaultValueFormat=0.000
# raven.measurement.basic.highFrequency.defaultValueFormat=0.000
#
raven.measurement.basic.deltaFrequency.defaultVal
ueFormat=0.000
308
# raven.measurement.basic.maxFrequency.defaultValueFormat=0.000
Amplitude To specify the default measurement display precision for amplitude
# raven.measurement.basic.amplitude.defaultValueFormat=0.000
To override the default measurement precision for amplitude
measurements and specify a different precision for Min Amplitude, Max
Amplitude, Peak Amplitude, or RMS Amplitude, uncomment and
edit the appropriate one of the following entries:
# raven.measurement.basic.minAmplitude.defaultValueFormat=0.000
# raven.measurement.basic.maxAmplitude.defaultValueFormat=0.000
# raven.measurement.basic.peakAmplitude.defaultValueFormat=0.000
# raven.measurement.basic.rmsAmplitude.defaultValueFormat=0.000
Power To specify the default measurement display precision for power
# raven.measurement.basic.power.defaultValueFormat=0.000
Overwrite behavior for recorders
You can specify a preference for what Raven should do when recording to
files if a new sound file has the same name as an existing file in the
selected recording directory, by editing the following entry:
raven.ui.audioRecorder.overwriteExistingFiles=false
When this preference is set to false (the default when Raven is installed),
Raven will create a new directory for the new file, rather than overwriting
the existing file as described in “File name collisions” in Chapter 4 (page
91). If you change the preference to true, the new file will replace the old
file without any warning.
The Memory Manager
You can use Raven’s Memory Manager window (Figure 11.8) to see how
much memory is used for each view of each file that is open, and to release
memory by deleting selected views or sound windows. You can open the
Memory Manager at any time by choosing Window > Memory Manager.
309
Figure 11.8. Memory manager
Figure 11.8. The Memory Manager window. When the window is first
opened, the tree view shown is fully collapsed. In this example, the
window would initially display only Sound 1. To expand or collapse an
item in the view, click on the opener icon to its left or double-click the
item’s name.
About Raven Raven runs within a simulated computer called the Java Virtual Machine
memory allocation (JVM). At any given time when Raven is running, a certain amount of
your computer’s memory, known as the heap, is allocated by the JVM for
use by Raven.1 The heap represents the memory that is available for
Raven’s use; the amount of memory that Raven is actually using at any
given moment may be less than the heap size. The heap can grow or
shrink as Raven’s memory needs change. Initially, when you first launch
Raven, the JVM allocates a relatively small heap, enough to display the
Raven window with no files open. As you open files, create views, and
perform other operations that require memory, more memory is allocated
to the heap as needed by the JVM. When you delete views or close sound
windows, the memory that was used for those objects becomes available
to Raven for re-use.
By default, the maximum heap size that the JVM can allocate is 256
megabytes. If Raven requests additional memory (e.g., to open another
sound file or to add a view to an existing sound window) when the heap
has already grown to its maximum size, an out-of-memory error will
1. The actual amount of memory available to Raven is slightly less than the heap
size, since part of the heap is used by the JVM itself.
310
occur. An out-of-memory error can also occur with a smaller heap size, if
there is not enough memory available in your computer for the JVM to
allocate the maximum heap size. This can happen if your computer does
not have enough physical memory installed, or if large amounts of the
installed memory are in use by other programs. When an out-of-memory
error occurs, Raven will ask if you want to use the Memory Manager to
release some memory.
The Memory Memory usage summary
Manager window The top portion of the Memory Manager window (Figure 11.8) displays
summary information about Raven’s current memory usage.
• Maximum Heap Size is the largest amount of memory that the JVM will
attempt to allocate to the heap.
• Current Heap Size is the current size of the heap that has been allocated
by the JVM to Raven. Each time Raven needs more memory than it has
available, the Current Heap Size increases, as the JVM allocates
successive increments of memory to the heap.
• Memory Used is the amount of the current heap that Raven is actually
using.
• Memory Available is the amount of Raven’s current heap that is not
presently in use. Memory Available is equal to Current Heap Size minus
Memory Used.
Configuring Maximum Heap Size
You can change the maximum heap size by clicking the Configure... button
in the Memory Manager window (Figure 11.8), and entering a new
maximum heap size in the Configure Maximum Heap Size dialog box that
appears. In order for the new maximum heap size to take effect, you must
restart Raven. If you do not restart immediately, the change will take effect
the next time you start Raven.
The Maximum Heap Size shown in the Memory Manager is
the largest value that the JVM will attempt to allocate to the
heap. The largest heap size that actually can be allocated
may be less than the Maximum Heap Size shown if there is
not enough memory available that is not being used by other
programs (including the operating system), or if there is insufficient physical memory installed in the computer A Maximum Heap Size of over 1.5 GB is only available if Raven was
installed with the 64-bit installer.
Resetting the maximum heap size from outside Raven
Take care to set the maximum heap size to a realistic value. If the value is
set beyond the range of the computer’s physical memory, Raven may fail
to start. If this happens, you will need to find the file that the Memory
311
Manager updated when you set the heap size, and change the associated
value. Once this value has been changed, Raven should restart normally.
For Mac OS, the file to change is Info.plist within the Raven application.
First, locate the Raven application within the Raven 1.4 folder and rightclick (or Control-click) on the Raven icon. Selecting “Show Package
Contents” from the menu should cause a folder named “Contents” to
appear. Within this folder is a file called Info.plist. Edit this file using the
text editor, and search within the file for -Xmx. Change the entry from
-Xmx[current value]M to -Xmx512M.
In Windows, the file to change is Raven.ini. This file is saved in the folder
in which you installed Raven. The default location is c:\Program
Files\Raven Pro 1.4\Raven.ini. Open this file for editing in a text editor
and change the value -Xmx[current value]M to -Xmx512M.
The memory usage list
The bottom portion of the Memory Manager window (Figure 11.8)
displays a list of all of the sound files presently open, and the size of each
one in megabytes (1 megabyte (Mbyte) = 1024 kilobytes; 1 kilobyte (Kbyte)
= 1024 bytes). To the left of the name of each sound (e.g., “Sound 1”,
“Sound 2”, etc.) is an opener icon; the shape of the opener icon depends on
the operating system under which you’re running Raven. Clicking on the
opener icon or double-clicking on the entry name expands the entry for
the sound. The expanded entry contains an entry for the sound samples
that constitute the signal, and one entry for each view that exists for that
sound (e.g., “Waveform 1”, “Spectrogram 1”, etc.). Each view entry again
has an associated opener icon. Clicking the opener icon or double-clicking
the entry name for a view expands the view’s entry to show an entry for
the view’s rendering data, and in the case of spectrogram and spectrogram
slice views, the spectrogram data. Rendering data represent the graphic
image displayed in a view’s pane within the signal window, and thus
depend in part on the size of a view’s pane. Changing the size of a view
pane or the entire window will change the size of the associated rendering
data, even though the sizes of the sound samples and spectrograms do not
change.
The memory usage list only tracks the typically large memory allocations
associated with sound windows and views. During the course of normal
operations, Raven uses memory for many other (relatively small) internal
data objects as well. Consequently, the Memory Used value is always
somewhat greater than the sum of the sizes shown in the list. Spectrogram
caching may further contribute to a discrepancy between the Memory
Used value and the sum of the items in the memory usage list.
Spectrogram caching occurs when you have requested multiple
spectrograms with identical parameters (e.g., for a spectrogram view and
a spectrogram slice view or for two spectrogram views that you view at
different scales). In these cases, Raven saves time and memory by
computing and storing the spectrogram data only once. Even though there
312
is only one copy of the cached spectrogram, the Memory Manager will
show a duplicate size entry for each view that uses the cached data.
You can delete a view or close a sound window by selecting its entry in the
Memory Manager list, then clicking Remove.
Reclaiming unused memory
When you delete a view or close a window, the memory that was used for
that object eventually becomes available for re-use (i.e., it is removed from
the Memory Used Category and added to Memory Available). This occurs
eventually rather than immediately because Raven only reclaims unused
memory when it needs to. You can force Raven to reclaim unused memory
at any time by clicking the Reclaim button in the Memory Manager
window. Clicking Reclaim will force Raven to display the most up-to-date
possible data on memory usage.
Raven program and documentation updates
Automatic updates From time to time, the Raven development team releases software and
documentation updates that fix errors in Raven or the Raven User’s
Manual. By default, Raven will automatically attempt to contact the Raven
update server via the Internet to see if any new update files are available
each time you start the program. To disable this behavior, edit the
following preferences entry, changing its value to false:
raven.updates.automatic=true
Installing updates If updates are found, Raven will display a dialog showing how many
update files are available, and their total size (Figure 11.9). You can then
choose whether or not to download the updates. Separate dialogs are
shown for program files and documentation files. If you choose not to
download update files, you can always download them later. You can
check for updates at any time by choosing Help > Check for Updates.
Figure 11.9. Update dialog
Figure 11.9. The download update files dialog.
313
There is no separate installation procedure required once the update files
have been downloaded, but you need to restart Raven for the updates to
take effect. If you do not restart immediately, the updates take effect the
next time you start Raven.
WINDOWS
On Windows computers, after updates are successfully
downloaded, Raven will ask if you want to restart
Raven immediately. If you click Yes, Raven automatically quits and restarts itself.
Troubleshooting Raven
Online resources Several online resources can help you if you run into a problem with
Raven. The first place to check is the Raven Pro FAQ, which lists common
problems and how to solve them. It can be found at http://
help.RavenSoundSoftware.com/forum/faq.php.
Another helpful place to look is the Raven Help Forum, where you can see
if other users have encountered your problem, and ask the Raven support
team for help. The forum also contains a list of How To’s for common
Raven tasks. It can be found at http://help.RavenSoundSoftware.com/
forum.
Finally, you can contact the Raven Team directly from Raven, by choosing
Help > Email Feedback. For more information, see “Contacting the Raven
development and support team” on page 315.
Reporting a bug If you find a bug in Raven, we’d like to hear about it. If possible, send your
bug report directly from Raven (Help > Email Feedback), because the
program will then send us additional information that may be helpful in
determining how to fix the bug.
If you are running Raven on Windows or Linux, you can get additional
information about your bug by running the Raven Console. In your Raven
folder, double-click on the icon labeled RavenConsole. Raven will launch
normally, with the addition of a text window called the console. If you can
reproduce your bug while running the console, additional error messages
may be displayed in the console window. Please send us these messages
along with your bug report.
If you are running Raven on a Mac, you can access the Raven Console by
opening /Applications/Utilities/Console. Please copy the contents of that
file to your bug report.
Performance tips If Raven is running slowly or stalling when you perform large operations,
there are several ways to speed it up. First, it is important to run Raven
with sufficient memory. By default, Raven runs with a maximum memory
allocation size of 256 megabytes. If your computer has more RAM than
314
this, you may wish to increase the amount Raven can use. To do this, open
the Memory Manager (Window > Memory Manager), then click on the
Configure button. From there, you can increase your maximum memory
allocation from 256 megabytes to a larger value, such as 512 megabytes.
The next time Raven is launched, it will use the new value for the
maximum memory allocation size.
Real-time signal displays can tax a computer’s CPU, especially on slower
computers. If necessary, you can reduce the load placed on your CPU by
these displays by making them smaller and/or hiding their axes.
If you’re working with a large number of selections, creating or
manipulating a selection can make Raven run slowly. You can fix that by
adding the following line to RavenPreferences.txt:
raven.ui.selections.drawOnlyActiveWhileModifyingWithMouse=true
This will make Raven run faster by making your other selections
disappear momentarily whenever a selection is created or modified. They
will reappear when you release your mouse button.
Another way to make Raven run faster while working with selections is to
turn off the “Selection Fill” checkbox in the Layout tab of the side panel.
This will remove the shading that fills selections, which will reduce the
load on your CPU.
While a detector is running, other operations in Raven may become slow
or unresponsive. To prevent this, minimize the sound window in which
the detector is running; then Raven’s operations should return to normal
speed while the detector continues to run in the background.
Contacting the
Raven
development and
support team
Raven has an online help and discussion forum where users can address
each other or the Raven team by asking questions, answering posted
questions, reporting bugs, or submitting feature requests. Visit us at http:/
/help.RavenSoundSoftware.com/ to view the forum.
The Raven developers want to hear from you. You can report a bug,
request support, or suggest a feature by choosing Help > Email Feedback,
and sending email from within Raven (Figure 11.10). Before clicking Send,
be sure to check that your correct email address appears in the From field,
so we can get back to you.2
2. After you fill in your email the first time, Raven will default to this address in
the future. To configure Raven to fill in your email address automatically if
you haven’t already used the function, open the RavenPreference.txt file using
any text editor, and fill in your email address in the line
raven.feedback.fromAddress=yourEmailAddress@yourISP
Save the edited file as plain text. The next time you launch Raven and use the
email feedback form, your address will be supplied automatically in the From
field.
315
Bug reporting Although you can of course use your usual email program to contact us (at
[email protected] for bugs and feature requests, or
[email protected] for support or other comments), there is
one situation in which you should use Raven’s email feedback feature, and
not another program. If Raven encounters an unrecoverable error, it will
ask if you want to submit a bug report. If you report the bug using Raven’s
email feedback form, Raven will include information along with your
message that can help us find the cause of the problem, and correct it in
future versions. (You can see the additional information that Raven
includes if you look in the Context tab.) If you send your message from
another program, this information will not be included.
Figure 11.10. Feedback form.
Figure 11.10. The Raven email feedback dialog.
Outgoing mail Depending on how your computer is configured, you may need to supply
server Raven with the address of your outgoing mail server, known as an SMTP
(Simple Mail Transfer Protocol) server. If you have trouble sending email
feedback from within Raven, remove the ‘#’ from the line
# raven.feedback.smtpServer=
316
and fill in the address of your SMTP server at the end of the line. If you
don’t know the address of your SMTP server, you may be able to find it in
a configuration or settings dialog of your usual email program, or you can
obtain it from your network administrator or internet service provider.
Problems sending If for some reason you are unable to send feedback when you want to do
feedback so (for example, because your computer is not connected to the Internet at
the time, or there is an unexpected problem transmitting the message),
you can choose File > Save As... to save a feedback file to be sent later. To
send the feedback file later, do one of the following:
•Choose File > Open in the Raven Email Feedback dialog box to open the
saved feedback file, then click Send.
•Send a message from your regular email program to
[email protected] or [email protected], and
include the saved feedback file as an attachment.
Again, if you have problems sending email feedback with
Raven that you cannot resolve, contact us with your own
email program or visit our help and discussion forum online at
http://help.RavenSoundSoftware.com/ If you do visit our
forum, please attach the feedback file to your post.
317
318
Appendix A
Digital Representation of Sound
About this appendix
This appendix provides a brief explanation of how sound is represented
digitally. An understanding of the basic principles introduced here will be
helpful in using Raven.
Digital sampling
Before a continuous, time-varying signal such as sound can be
manipulated or analyzed with a digital computer, the signal must be
acquired or digitized by a hardware device called an analog-to-digital (A/D)
converter, or digitizer. The digitizer repeatedly measures or samples the
instantaneous voltage amplitude of a continuously varying (analog) input
signal at a particular sample rate, typically thousands or tens of thousands
of times per second (Figure A.1). In the case of an audio signal, this timevarying voltage is proportional to the sound pressure at a device such as a
microphone. The digital representation of a signal created by the digitizer
thus consists of a sequence of numeric values representing the amplitude
of the original waveform at discrete, evenly spaced points in time.
319
Appendix A: Digital Representation of Sound
Figure A.1. Digital sampling
Figure A.1. Sampling to create digital representation of a pure tone
signal. The blue sinusoidal curve represents the continuous analog
waveform being sampled. Measurements of the instantaneous
amplitude of the signal are taken at a sample rate of 1/Δt. The resulting
sequence of amplitude values is the digitized signal.
The precision with which the digitized signal represents the continuous
signal depends on two parameters of the digitizing process: the rate at
which amplitude measurements are made (the sample rate or sampling
frequency), and the number of bits used to represent each amplitude
measurement (the sample size or bit depth).
Sample rate
Raven’s Configure Recorder dialog box enables you to choose the sample
rate at which a signal is to be digitized. The choices available are
determined by the digitizer hardware and the program (called an audio
input plug-in in Raven) that controls the digitizer; most digitizers have two
or more sample rates available. Commercial digital audio applications use
sample rates of 44.1 kHz (for audio compact discs) or 48 kHz (for digital
audio tape). Once a signal is digitized, its sample rate is fixed.
In order to interpret a sequence of numbers as representing a time-varying
signal, one needs to know the sample rate. Thus, when a digitized signal is
saved in a file format that is designed for saving sound information (such
as AIFF or WAVE), information about the sample rate is saved along with
the actual data points comprising the signal.
Aliasing and the The more frequently a signal is sampled, the more precisely the digitized
Nyquist frequency signal represents temporal changes in the amplitude of the original signal.
The sample rate that is required to make an acceptable representation of a
waveform depends on the signal’s frequency. More specifically, the
320
sample rate must be more than twice as high as the highest frequency
contained in the signal. Otherwise, the digitized signal will have
frequencies represented in it that were not actually present in the original
at all. This appearance of phantom frequencies as an artifact of inadequate
sample rate is called aliasing (Figure A.2).
Figure A.2. Aliasing (sampled waveform)
Figure A.2. Aliasing as a result of inadequate sample rate. Vertical
lines indicate times at which samples are taken. (a) A 500 Hz pure tone
sampled at 8000 Hz. The blue sinusoidal curve represents the
continuous analog waveform being sampled. There are 16 sample
points (= 8000/500) in each cycle of the waveform. If the same analog
signal were sampled at 800 Hz (red sample points), there would be
fewer than two points per cycle, and aliasing would result. (b) The
aliased waveform that would be represented by sampling the 500 Hz
signal at a sample rate of 800 Hz (Nyquist frequency = 400 Hz). Since
the original waveform was 100 Hz higher than the Nyquist frequency,
the aliased signal is 100 Hz below the Nyquist frequency, or 300 Hz.
The highest frequency that can be represented in a digitized signal
without aliasing is called the Nyquist frequency, and is equal to half the
frequency at which the signal was digitized. The highest frequency shown
in a spectrogram or spectrum calculated by Raven is always the Nyquist
frequency of the digitized signal. If the only energy above the Nyquist
frequency in the analog signal is in the form of low-level, broadband
noise, the effect of aliasing is to increase the noise in the digitized signal.
However, if the spectrum of the analog signal contains any peaks above
the Nyquist frequency, the spectrum of the digitized signal will contain
spurious peaks below the Nyquist frequency as a result of aliasing. In
321
spectrograms, aliasing is recognizable by the appearance of one or more
inverted replicates of the real signal, offset in frequency from the original
(Figure A.3).
Figure A.3. Spectrogram of aliased signal.
Figure A.3. Appearance of aliasing in spectrogram views. (a)
Spectrogram of a bearded seal song signal digitized at 11025 Hz. All of
the energy in the signal is below the Nyquist frequency (5512.5 Hz);
only the lowest 2300 Hz is shown. The red line is at 1103 Hz, one-fifth
of the Nyquist frequency. (b) The same signal sampled at 2205 Hz
(one-fifth of the original rate; Nyquist frequency, 1102.5 Hz) without an
anti-aliasing filter. The frequency downsweep in the first ten seconds of
the original signal appears in inverted form in this undersampled signal,
due to aliasing. (c) The same signal as in (b), but this time passed
through a low-pass (anti-aliasing) filter with a cutoff of 1100 Hz before
being digitized. The downsweep in the first ten seconds of the original
signal, which exceeds the Nyquist frequency, does not appear because
it was blocked by the filter.
The usual way to prevent aliasing is to pass the analog signal through a
low-pass filter (called an anti-aliasing filter) before digitizing it, to remove
any energy at frequencies greater than the Nyquist frequency. (If the
original signal contains no energy at frequencies above the Nyquist
322
frequency or if it contains only low-level broadband noise, this step is
unnecessary.)
Sample size (amplitude resolution)
The precision with which a sample represents the actual amplitude of the
waveform at the instant the sample is taken depends on the sample size or
number of bits (also called bit depth) used in the binary representation of
the amplitude value. Some digitizers can take samples of one size only;
others allow you to choose (usually through software) between two or
more sample sizes. Raven’s default audio input device plug-in allows you
to choose between 8-bit and 16-bit samples. An 8-bit sample can resolve
256 (=28) different amplitude values; a 16-bit converter can resolve 65,536
(=216) values. Sound recorded on audio CDs is stored as 16-bit samples.
When a sample is taken, the actual value is rounded to the nearest value
that can be represented by the number of bits in a sample.
Since the actual analog value of signal amplitude at the time of a sample is
usually not precisely equal to one of the discrete values that can be
represented exactly by a sample, there is some digitizing error inherent in
the acquisition process (Figure A.4), which results in quantization noise in
the digitized signal. The more bits used for each sample, the less
quantization noise is contained in the digitized signal. If you listen to a
signal digitized with 8-bit samples using high-quality headphones, you
can hear the quantization noise as a low-amplitude broadband hiss
throughout the recording. Signals digitized with 16-bit samples typically
have no detectable hiss. The ratio between the value of the highest
amplitude sample that can be represented with a given sample size and
the lowest non-zero amplitude is called the dynamic range of the signal,
and is usually expressed in decibels (dB). The dynamic range corresponds
to the ratio in amplitude between the loudest sound that can be recorded
and the quantization noise. The dynamic range of a digitized sound is 6
dB/bit.1
1. The dynamic range of a signal in decibels is equal to 20 log(Amax/Amin),
where Amax and Amin are the maximum and minimum amplitude values
in the signal. For a digitized signal, Amax/Amin = 2n, where n is the number
of bits per sample. Since log(2n) = 0.3n, the dynamic range of a signal is 6
dB/bit.
323
Figure A.4. Digitizing error
Figure A.4. Digitizing error with a hypothetical 2-bit sample size. 2-bit
samples can represent only four different amplitude levels. The blue
sinusoidal curve represents the continuous analog waveform being
sampled. At each sample time (vertical lines), the actual amplitude
levels are rounded to the nearest value that can be represented by a 2bit sample (horizontal lines). The amplitude values stored for most
samples (dots) are slightly different from the true amplitude level of the
signal at the time the sample was taken.
Specifying sample
sizes when
acquiring and
saving signals
Raven lets you specify the sample size for a signal when you first acquire
it, and again when you save the signal to a file. The set of sample sizes that
are available during acquisition is determined by the sound input plug-in
that you select (sound acquisition is discussed in Chapter 4, “Signal
Acquisition (Recording)”). While Raven is actually working with a signal,
samples are always represented by 32-bit floating-point values. When you
save a signal with a sample size other than the sample size that the signal
had when it was acquired or opened, Raven scales the values to the
sample size that you select when you specify the format of the file (saving
files is discussed in Chapter 1, “Getting Started”). For example, if you
open a file containing 8-bit samples, and then save the signal with 16-bit
samples, each sample value will be multiplied by 28. This scaling ensures
that a full-scale value in the original signal is still a full-scale value in the
saved signal, even if the sample size differ.
Although you can acquire a signal with 8 bits and then save it
with a larger sample size, the saved signal will retain the
smaller dynamic range (and audible quantization hiss) of the
8-bit signal. This is because the quantization noise is scaled
along with the desired signal when 8-bit signals are scaled to
the larger sample size.
324
Storage requirements
The increased frequency bandwidth obtainable with higher sample rates
and the increased dynamic range obtainable with larger samples both
come at the expense of the amount of memory required to store a digitized
signal. The minimum amount of storage (in bytes) required for a digitized
signal is the product of the sample rate (in samples/sec), the sample size
(in bytes; one byte equals 8 bits), and the signal duration (seconds). Thus, a
10-second signal sampled at 44.1 kHz with 16-bit (2-byte) precision
requires 882,000 bytes (= 10 sec x 44,100 samples/sec x 2 bytes/sample), or
about 861 Kbytes of storage (1 Kbyte = 1024 bytes). The actual amount of
storage required for a signal may exceed this minimum, depending on the
format in which the samples are stored.
The amount of time that it takes Raven to calculate a spectrogram of a
signal depends directly on the number of samples in that signal. Thus,
spectrograms take longer to calculate for signals digitized at higher rates.
However, the sample size at which a signal is acquired or saved does not
affect the speed of spectrogram calculation, because Raven always
converts signals to a 16-bit representation for internal operations, even if
the signal was initially acquired or saved with a different sample size.
325
326
Appendix B
A Biologist’s Introduction to
Spectrum Analysis
About this appendix
This appendix provides some conceptual background for making and
interpreting spectrogram and spectrogram slice views with Raven. It
introduces the short-time Fourier transform (STFT), the mathematical
technique used by Raven for making spectrograms. We do not discuss the
mathematics of the STFT, but instead treat it here as a black box. This
black box has controls on its outside that affect its operation in important
ways. One aim of this appendix is to convey enough qualitative
understanding of the behavior of this box to allow intelligent use of its
controls, without delving into the box’s internal mechanism. Specific
details of the controls are covered in Chapter 3, “Spectrographic
Analysis”. A second aim of this appendix is to explain some of the
limitations and tradeoffs intrinsic to spectrum analysis of time-varying
signals. More rigorous mathematical treatments of spectral analysis, at
several levels of sophistication, can be found in the references listed at the
end of the appendix.
Several approaches can be taken to explaining the fundamentals of digital
spectrum analysis. The approach taken in this appendix is geared
specifically to spectrum analysis with Raven; thus some of the terms and
concepts used here may not appear in other, more general discussions of
spectrum analysis, such as those listed at the end of the appendix.
The discussions in this appendix assume a basic understanding of how
sound is recorded and represented digitally. If you are not already
acquainted with concepts such as sample rate and sample size, you should
read Appendix A, “Digital Representation of Sound” before proceeding.
What sound is
Sound consists of traveling waves of alternating compression and
rarefaction in an elastic medium (such as air or water), generated by some
vibrating object (a sound source).
Sound pressure is the (usually small) alternating incremental change in
pressure from ambient pressure that results from a sound. When no sound
is present in a medium (i.e., there is no propagating pressure change), we
say that sound pressure is zero, even though the medium does exert some
static ambient pressure. The dimensions of pressure are force per unit
327
Appendix B: A Biologist’s Introduction to Spectrum Analysis
area. The usual unit of sound pressure is the pascal (abbreviated Pa); one
pascal equals one newton per square meter. Since the smallest audible
sound pressures in air are on the order of 10-6 Pa, sound pressures are
usually expressed in µPa.
To measure or record sound at a particular location in space, we use a
device such as a microphone that responds to sound pressure. A
microphone produces a time-varying electrical voltage that is
proportional to the increase or decrease in local pressure that constitutes
sound. This continuous time-varying voltage is an electric analog of the
acoustic signal. The continuous electric signal can be converted to a digital
representation suitable for manipulation by a computer as discussed in
Appendix A, “Digital Representation of Sound”.
Time domain and frequency domain representations of sound
Any acoustic signal can be graphically or mathematically depicted in
either of two forms, called the time domain and frequency domain
representations. In the time domain, instantaneous pressure is represented
as a function of time. Figure B.1a shows the time domain representation of
the simplest type of acoustic signal, a pure tone. Such a signal is called a
sinusoid because its amplitude is a sine function of time, characterized by
some frequency, which is measured in cycles per second, or Hertz (Hz).
The frequency of a sinusoid is most easily determined by measuring the
length of one period, which is the reciprocal of the frequency. The
amplitude of the signal in the time domain is measured in pressure units.
(Once an acoustic signal has been converted by a microphone into an
electrical signal, amplitude is measured as voltage, which is directly
proportional to the sound pressure.) In the frequency domain, the
amplitude of a signal is represented as a function of frequency. The
frequency domain representation of a pure tone is a vertical line (Figure
B.1b).
328
Figure B.1. Time and freq domain
Figure B.1. Time domain and frequency domain representations of an
infinitely long pure sinusoidal signal. (a) Time domain. t is the period of
the sinusoid. (b) Frequency domain. f is the frequency of the sinusoid.
Any sound, no matter how complex, can be represented as the sum of a
series of pure tones (sinusoidal components). Each tone in the series has a
particular amplitude, and a particular phase relationship (i.e., it may be
shifted in time) relative to the others. The frequency composition of
complex signals is usually not apparent from inspection of the time
domain representation. Spectrum analysis is the process of converting the
time domain representation of a signal to a frequency domain
representation that shows how different frequency components contribute
to the sound. Frequency domain representations of sounds are often more
intuitively interpretable because the mammalian auditory system
(specifically the cochlea) performs a type of spectrum analysis in
converting vibrations of the eardrum into neural impulses. Our auditory
perception is thus based on a frequency domain representation of sounds.
The complete frequency domain representation of a signal consists of two
parts. The magnitude spectrum (Figure B.2b) contains information about the
relative magnitude of each frequency component in the entire signal. The
phase spectrum (Figure B.2c) contains information about the relative phase
or timing relationships among the frequency components. Since the phase
spectrum is rarely of practical use in most bioacoustic work and is not
provided by Raven, it is not discussed further here. Henceforth, unless
329
otherwise noted, we use the term “spectrum” to refer to the magnitude
spectrum alone.
Figure B.2. Time, freq domains-- 2 tones
Figure B.2. Time domain and frequency domain representations of an
infinitely long sound consisting of two tones, with frequencies of 490 Hz
and 800 Hz. (a) Time domain. (b) Magnitude spectrum in frequency
domain. (c) Phase spectrum in frequency domain. The phase of the
frequency component at 500 Hz is arbitrarily taken as a reference and
assigned a phase value of 0.
The Fourier transform is a mathematical function that converts the time
domain form of a signal (which is the representation directly produced by
most measuring and recording devices) to a frequency domain
representation, or spectrum. When the signal and spectrum are
represented as a sequence of discrete digital values, a version of the
Fourier transform called the discrete Fourier transform (DFT) is used. The
330
input to the DFT is a finite sequence of values— the amplitude values of
the signal— sampled (digitized) at regular intervals. The output is a
sequence of values specifying the amplitudes of a sequence of discrete
frequency components, evenly spaced from 0 Hz to half the sampling
frequency (Figure B.3). Raven implements the DFT using an algorithm
known as the fast Fourier transform (FFT).
Figure B.3. DFT schematic
Figure B.3. Schematic representation of the discrete Fourier transform
(DFT) as a black box. The input to the DFT is a sequence of digitized
amplitude values (x0, x1, x2, ... xN-1) at N discrete points in time. The
number of input values N is called the DFT size. The output is a
sequence of amplitude values (A0, A1, A2, ... A(N/2)) at N/2 discrete
frequencies. The highest frequency, f(N/2)-1, is equal to half the sample
rate (= 1/(2T), where T is the sampling period, as shown in the figure).
The output can be plotted as a magnitude spectrum.
In practice, a spectrum is always made over some finite time interval. This
interval may encompass the full length of a signal, or it may consist of
some shorter part of a signal.
Spectral analysis of time-varying signals: spectrograms and STFT analysis
Most signals of biological interest change over time in frequency (spectral)
composition. Indeed the changes in spectrum over time are often among
the most interesting aspects of such signals. But in order to create a
spectrum, we must examine an interval of time— there is no way to
measure a signal’s “instantaneous” spectrum. An individual magnitude
spectrum of a signal provides no information about temporal changes in
frequency composition during the interval over which the spectrum is
made. If we were to make a single magnitude spectrum over the entire
duration of a spectrally varying signal such as a typical bird song, we
331
would have a representation of the relative intensities of the various
frequency components of the signal, but we would have no information
about how the intensities of different frequencies varied over time during
the signal.
To see how the frequency composition of a signal changes over time, we
can examine a sound spectrogram.1 The spectrograms produced by Raven
plot frequency on the vertical axis versus time on the horizontal; the
amplitude of a given frequency component at a given time is represented
by a color (by default, grayscale) value (Figure B.4).
Figure B.4. Spectrogram example.
Figure B.4. Smoothed sound spectrogram of part of a song of a
chestnut-sided warbler, digitized at 44.1 kHz.
Spectrograms are produced by a procedure known as the short-time Fourier
transform (STFT). The STFT divides the entire signal into a series of
successive short time segments, called records (or frames). Each record is
used as the input to a DFT, generating a series of spectra (one for each
record). To display a spectrogram, the spectra of successive records are
plotted side by side with frequency running vertically and amplitude at
each frequency represented by a color (by default, grayscale) value.
Raven’s spectrogram slice view displays the spectrum of one record at a
time as a line graph, with frequency on the horizontal axis, and amplitude
on the vertical axis. A spectrogram can be characterized by its DFT size,
expressed as the number of digitized amplitude samples that are
processed to create each individual spectrum.
The STFT can be considered as equivalent in function to a bank of N/2 + 1
bandpass filters, where N is the DFT size. Each filter is centered at a
slightly different analysis frequency. The output amplitude of each filter is
proportional to the amplitude of the signal in a discrete frequency band or
bin, centered on the analysis frequency of the filter. In this “filterbank”
model of STFT analysis, the spectrogram is considered as representing the
1. Sound spectrograms are sometimes called sonagrams. Strictly speaking, however, the term sonagram is a trademark for a sound spectrogram produced by
a particular type of spectrum analysis machine called a Sonagraph, produced
by the Kay Elemetrics Co.
332
time-varying output amplitudes of filters at successive analysis
frequencies plotted above each other, with amplitude again represented
by color (by default, grayscale) values. A spectrogram can be
characterized by its bandwidth, the range of input frequencies around the
central analysis frequency that are passed by each filter. All of the filters in
a spectrogram have the same bandwidth, irrespective of analysis
frequency.
Record length, bandwidth, and the time-frequency uncertainty principle
The record length of a STFT determines the time analysis resolution (Δt) of
the spectrogram. Changes in the signal that occur within one record (e.g.,
the end of one sound and the beginning of another, or changes in
frequency) cannot be resolved as separate events. Thus, shorter record
lengths allow better time analysis resolution.
Similarly, the bandwidth of a STFT determines the frequency analysis
resolution (Δf) of the spectrogram: frequency components that differ by less
than one filter-bandwidth cannot be distinguished from each other in the
output of the filterbank. Thus a STFT with a relatively wide bandwidth
will have poorer frequency analysis resolution than one with a narrower
bandwidth.
Ideally we might like to have very fine time and frequency analysis
resolution in a spectrogram. These two demands are intrinsically
incompatible, however: the record length and filter bandwidth of a STFT
are inversely proportional to each other, and cannot be varied
independently. Although a short record length yields a spectrogram with
finer time analysis resolution, it also results in wide bandwidth filters and
correspondingly poor frequency analysis resolution. Thus a tradeoff exists
between how precisely a spectrogram can specify the spectral (frequency)
composition of a signal and how precisely it can specify the time at which
the signal exhibited that particular spectrum.
The relationship between record length and filter bandwidth applies to
each of the individual spectra that collectively constitute a spectrogram.
Figure B.5 illustrates the relationship between record length and filter
bandwidth in individual spectra. The two spectra, of a 2000 Hz pure tone
digitized at 22.05 kHz, were made with different record lengths and thus
different bandwidths. Spectrum (a), with a record length of 1024 points
(46.0 mS), shows a fairly sharp peak at 2000 Hz because of its relatively
narrow bandwidth (35.3 Hz) filter; spectrum (b), with a record length of
256 points (11.5 mS), corresponding to a wider bandwidth (141 Hz) filter,
has poorer frequency resolution.
333
Figure B.5. Window length - bandwidth relationship
Figure B.5. Relationship between record length and 3 dB bandwidth.
Each view is of a single spectrum of a 2000 Hz tone, digitized at 22.05
kHz. In both spectra, window function = Blackman. The highlighted
band in each spectrum shows the 3 dB bandwidth.
(a) Window size = 1024 points = 46.0 mS; 3 dB bandwidth = 35.3 Hz.
(b) Window size = 256 points = 11.6 mS; 3 dB bandwidth = 141 Hz.
Making spectrograms
A spectrogram produced by Raven is a two-dimensional grid of discrete
data points on a plane in which the axes are time and frequency.
Ordinarily, this grid is not apparent because by default Raven smooths the
spectrogram display, interpolating color values for pixels that are between
the gridpoints where values were calculated by the STFT algorithm. If you
turn off spectrogram smoothing and stretch the time and frequency scales
adequately, the discrete nature of the spectrogram becomes evident
(Figure B.6). In a spectrogram displayed with smoothing turned off, the
color of each box represents an estimate of the logarithm of the relative
sound power (in decibels) in a particular frequency band over a particular
time interval. The center point of the box is at the center of the
corresponding frequency band and time interval.
334
Figure B.6. Boxy spectrogram
Figure B.6. Same spectrogram as in Figure B.4, with smoothing turned
off. The grayscale value in each box represents an estimate of the
relative power in the corresponding frequency band and time interval.
Filter bandwidth = 124 Hz, window size (record length) = 512 samples
(= 11.6 mS). Grid spacing = 5.8 mS x 86.1 Hz.
Raven lets you specify the spacing between gridpoints in the time
dimension and thus the width of the boxes in an unsmoothed
spectrogram. In Raven’s Configure Spectrogram dialog, you can specify
the time grid spacing (also called hop size) directly, or indirectly by
specifying the amount of overlap between successive records. (You specify
the record length of a spectrogram in Raven by entering the size of a
window function. Window functions are discussed in “Window functions”
on page 342.) The spacing between gridpoints in the frequency dimension
is determined by the DFT size. Raven chooses DFT size automatically,
using the smallest power of 2 which is greater than or equal to the window
size (in samples).
The relationships between time grid spacing and record overlap, and
between frequency grid spacing and DFT size are discussed below. See
Chapter 5, “Spectrographic Analysis”, for a detailed discussion of how to
control these parameters in Raven.
Grid spacing should not be confused with analysis resolution. Analysis
resolution for time and frequency are determined by the record length and
bandwidth of a STFT, respectively. Analysis resolution describes the
amount of smearing or blurring of temporal and frequency structure at
each point on the grid, irrespective of the spacing between these points.
The following sections seek to clarify the concepts of analysis resolution
and grid spacing by showing examples of spectrograms that illustrate the
difference between the two.
Analysis resolution
and the timefrequency
uncertainty
principle
At each point on the spectrogram grid, the tradeoff between time and
frequency analysis resolution is determined by the relationship between
record length and bandwidth, as discussed above. According to the
uncertainty principle, a spectrogram can never have extremely fine
analysis resolution in both the frequency and time dimensions.
335
For example, Figure B.7 shows two spectrograms of the same signal that
differ in record length and hence, bandwidth. In spectrogram (a), with a
record length of 64 points (= 2.9 mS; bandwidth = 496 Hz), the beginning
and end of each tone can be clearly distinguished and are well-aligned
with the corresponding features of the waveform. However, the frequency
analysis resolution is poor: each tone appears as a bar that is nearly 1200
Hz in thickness. In spectrogram (c), the record length is 512 points, or 23
mS (filter bandwidth = 61.9 Hz), or about as long as each tone in the signal.
Most of the records therefore span more than one tone, in some cases
including a tone and a silent interval, in other cases including two tones
and an interval. The result is poor time resolution: the beginning and end
of the bars representing the tones are fuzzy and poorly aligned with
features of the waveform (compare, for example, the beginning time of the
first pulse in the waveform with the corresponding bar in the
spectrogram). However, this spectrogram has much better frequency
resolution than spectrogram (a): the bar representing each tone is only
about 100 Hz in thickness.
336
Figure B.7. Time vs freq resolution.
Figure B.7. Effect of record length and filter bandwidth on time and
frequency resolution. The signal consists of a sequence of four tones
with frequencies of 1, 2, 3, and 4 kHz, at a sample rate of 22.05 kHz.
Each tone is 20 mS in duration. The interval between tones is 10 mS.
Both spectrograms have the same time grid spacing = 1.45 mS, and
window function = Hann. The selection boundaries show the start and
end of the second tone.
(a) Wide-band spectrogram: record length = 64 points ( = 2.90 mS), 3
dB bandwidth = 496 Hz.
(b) Waveform, showing timing of the tones.
(c) Narrow-band spectrogram: record length = 512 points ( = 23.2 mS),
3 dB bandwidth = 61.9 Hz.
The waveform between the spectrograms shows the timing of the
pulses.
What is the “best” window size to choose? The answer depends on how
rapidly the signal’s frequency spectrum changes, and on what type of
information is most important to show in the spectrogram, given your
particular application. For many applications, Raven’s default window
size (512 samples) provides a reasonable balance between time and
frequency resolution. If you need to observe very short events or rapid
changes in the signal, a shorter window may be better; if precise frequency
representation is more important, a longer window may be better2. If you
need better time and frequency resolution than you can achieve in one
2. If the features that you’re interested in are distinguishable in the waveform
(e.g., the beginning or end of a sound, or some other rapid change in amplitude), you’ll achieve better precision and accuracy by making time measurements on the waveform rather than the spectrogram.
337
spectrogram, you may need to make two spectrograms: a wide-band
spectrogram with a small window for making precise time measurements,
and a narrow-band spectrogram with a larger window for precise
frequency measurements.
Time grid spacing Time grid spacing (also called hop size) is the time between the beginnings
and window overlap of successive records. In an unsmoothed spectrogram, this interval is
visible as the width of the individual boxes (Figure B.6). Successive
records that are analyzed may be overlapping (positive overlap),
contiguous (zero overlap), or discontiguous (negative overlap). Overlap
between records is usually expressed as a percentage of the record length.
Figure B.8 illustrates the different effects of changes to record length and
time grid spacing. The signal is a frequency-modulated tone that sweeps
upward in frequency from 4 to 6 kHz, sampled at 22.05 kHz.
Spectrograms (a) and (c) both have a record length of 512 points (= 23.2
mS; 3 dB bandwidth = 61.9 Hz). (a) was made with 0% overlap (time grid
spacing = 23.2 mS), whereas (c) was made with an overlap of 93.8% (time
grid spacing = 1.45 mS). In the low-resolution spectrogram (a), each box is
as wide as one data record, which in turn is one quarter of the length of the
tone. The result is a spectrogram that gives an extremely misleading
picture of the signal. Spectrogram (c), with a greater record overlap, is
much “smoother” than the one with less overlap, and it more accurately
portrays the continuous frequency modulation of the signal. It still
provides poor time analysis resolution, however, because of its large
record length— notice the fuzzy beginning and end of the spectrogram
image of the tone and the poor alignment with the beginning and end of
the tone in the waveform. Comparison of the spectrograms in Figure B.8
demonstrates that improved time grid spacing is not a substitute for finer
time analysis resolution, which can be obtained only by using a shorter
record.
338
Figure B.8. Window size window size - overlap
Figure B.8. Different effects on spectrograms of changing record length
(= window size, or time analysis resolution) and time grid spacing. The
signal is a frequency-modulated tone, 100 mS long, sampled at 22.05
kHz. The tone sweeps upward in frequency from 4 to 6 kHz.
Spectrograms (a) and (c) have the same window size, but (c) has finer
time grid spacing (higher record overlap). (c) and (d) have the same
time grid spacing, but (d) has a shorter record length (finer time
analysis resolution).
(a) Record length = 512 points = 23.2 mS (3 dB bandwidth = 61.9 Hz);
Time grid spacing = 23.2 mS (overlap = 0%).
(b) Waveform view, with duration of tone highlighted.
(c) Record length = 512 points = 23.2 mS (3 dB bandwidth = 61.9 Hz);
Time grid spacing = 1.45 mS (overlap = 93.8%).
(d) Record length = 64 points = 2.9 mS (3 dB bandwidth = 448 Hz);
Time grid spacing = 1.45 mS (overlap = 50%).
Frequency grid Frequency grid spacing is the difference (in Hz) between the central
spacing and DFT analysis frequencies of adjacent filters in the filterbank modeled by a
size STFT, and thus the size of the frequency bins in a spectrogram. In an
unsmoothed spectrogram, this spacing is visible as the height of the
individual boxes (Figure B.6). Frequency grid spacing depends on the
sample rate (which is fixed for a given digitized signal) and DFT size. The
relationship is
frequency grid spacing = (sample rate) / DFT size
339
where frequency grid spacing and sample rate are measured in Hz, and
DFT size is measured in samples. Thus a larger DFT size draws the
spectrogram on a grid with finer frequency resolution (smaller frequency
bins, vertically smaller boxes). The number of frequency bins in a
spectrogram or spectrum is half the DFT size, plus one.
Recall that the DFT size is the number of samples processed to calculate
the spectrum of a record. Thus the DFT size would ordinarily be equal to
the record length. However, Raven’s DFT algorithm requires that the size
of the DFT be a power of 2. Therefore Raven automatically chooses the
smallest DFT size that is a power of 2 greater than or equal to the record
size. The sample data in each record are then filled out with zeros (“zeropadded”) to make the record length the same as the chosen DFT size. Zero
padding provides the right number of samples to match the chosen DFT
size without altering the spectrum of the data.
Spectral smearing The spectra that constitute a spectrogram produced by a STFT are
and sidelobes “imperfect” in several respects. First, as discussed above, each filter
simulated by the STFT has a finite band of frequencies to which it
responds; the filter is unable to discriminate different frequencies within
this band. According to the uncertainty principle, the filter bandwidth can
be reduced— thus improving frequency resolution— only by analyzing a
longer record, which reduces temporal resolution.
Second, the passbands of adjacent filters overlap in frequency, so that
some frequencies are passed (though partially attenuated) by more than
one filter (Figure B.9). Consequently, when a spectrum or spectrogram is
constructed by plotting the output of all of the filters, a signal consisting of
a pure tone becomes “smeared” in frequency (Figure B.9d).
340
Figure B.9. Spectral smearing-- overlapping filters
Figure B.9. Spectral smearing resulting from overlapping bandpass
filters. (a) A single hypothetical bandpass filter centered at frequency f0.
When the input to the filter is a pure tone at frequency f0, the output
amplitude is A0. For clarity of illustration, sidelobes to the main
passband are not shown (see text and Figure B.10). (b) Two
overlapping filters, centered at frequencies f0 and f1. When the filter
centered at f1 is presented with the same input as in (a), its output
amplitude is A1. (c) A bank of overlapping filters simulated by a STFT.
Frequency f0 falls within the passbands of the filter centered at f0, and
of two filters (blue and green) on either side. (d) Spectrum of a pure
tone signal of frequency f0 produced by the filterbank shown in (c). The
spectrum consists of one amplitude value from each filter. Because the
filters overlap, the spectrum is smeared, showing energy at frequencies
adjacent to f0. The shape of the resulting spectrum is the same as that
of a single filter.
341
Third, each filter does not completely block the passage of all frequencies
outside of its nominal passband. For each filter there is an infinite series of
diminishing sidelobes in the filter’s response to frequencies above and
below the passband (Figure B.10). These sidelobes arise because of the
onset and termination of the portion of the signal that appears in a single
record. Since a spectrum of a pure tone made by passing the tone through
a set of bandpass filters resembles the frequency response of a single filter
(Figure B.9), a STFT spectrum of any signal (even a pure tone) contains
sidelobes.
Figure B.10. Filter sidelobes
Figure B.10. Frequency response of a hypothetical bandpass filter
from a set of filters simulated by a short-time Fourier transform,
showing sidelobes above and below the central lobe, or passband. The
magnitude of the sidelobes relative to the central lobe can be reduced
by use of a window function (see text). Note that a spectrum produced
by passing a pure tone through a set of overlapping filters is shaped like
the frequency response of a single one of the filters (see Figure B.9).
Window functions The magnitude of the sidelobes (relative to the magnitude of the central
lobe) in a spectrogram or spectrum of a pure tone is related to how
abruptly the windowed signal’s amplitude changes at the beginning and
end of a record. A sinusoidal tone that instantly rises to its full amplitude
at the beginning of a record, and then instantly falls to zero at the end, has
higher sidelobes than a tone that rises and falls gradually in amplitude
(Figure B.11).
342
Figure B.11. Windowing
Figure B.11. Relationship between abruptness of onset and
termination of signal in one record and spectral sidelobes. Each panel
shows a signal on the left, and its spectrum on the right.
(a) A single record of an untapered sinusoidal signal has a spectrum
that contains a band of energy around the central frequency, flanked by
sidelobes, as if the signal had been passed through a bank of
bandpass filters like the one shown in Figure B.10.
(b) A single record of a sinusoidal signal multiplied by a “taper” or
window function, has smaller sidelobes.
The magnitude of the sidelobes in a spectrum or spectrogram can be
reduced by multiplying the record by a window function that tapers the
waveform as shown in Figure B.11. Tapering the waveform in the record is
equivalent to changing the shape of the analysis filter (in particular,
lowering its sidelobes). Each window function reduces the height of the
highest sidelobe to some particular proportion of the height of the central
peak; this reduction in sidelobe magnitude is termed the sidelobe rejection,
and is expressed in decibels (Table B.1). Given a particular record length,
the choice of window function thus determines the sidelobe rejection, and
also the width of the center lobe. The width of the center lobe in the
spectrum of a pure tone is the filter bandwidth.
Table B.1. Sidelobe rejection for Raven’s five window
types. The sidelobe rejection for each type is expressed as
the height of the highest sidelobe relative to the peak of the
main lobe.
Window type
Sidelobe rejection (dB)
Blackman
-57
Hamming
-41
343
Table B.1. Sidelobe rejection for Raven’s five window
types. The sidelobe rejection for each type is expressed as
the height of the highest sidelobe relative to the peak of the
main lobe.
Window type
Sidelobe rejection (dB)
Hann
-31
Rectangular
-13
Triangular
-25
For further reading
The books and articles listed below can provide entry at several levels into
the vast literature on spectrum analysis and digital signal processing.
Beecher, M. D. (1988). Spectrographic analysis of animal vocalizations:
Implications of the “uncertainty principle.” Bioacoustics 1(1), 187207.
Includes a discussion of choosing an “optimum” filter bandwidth
for the analysis of frequency-modulated bioacoustic signals.
Bradbury, J. & Vehrencamp, S. (1998). Principles of animal communication.
Sinauer Associates: Sunderland, MA.
Chapter 3 provides an excellent introduction for non-specialist
readers to the principles of spectrum analysis, and also discusses
spectral properties of the basic types of animal acoustic signals.
Cohen, L. (1995). Time-frequency analysis. Prentice-Hall: Englewood Cliffs,
NJ.
Hlawatsch, F. & Boudreaux-Bartels, G. F. (1992). Linear and quadratic
time-frequency signal representations. IEEE Signal Processing
Magazine 9(2), 21-67.
A technical overview and comparison of the properties of a variety
of time-frequency representations (including spectrograms),
written for engineers.
Jaffe, D. A. (1987). Spectrum analysis tutorial. Part 1: The Discrete Fourier
Transform; Part 2: Properties and applications of the Discrete
Fourier Transform. Computer Music Journal 11(3), 9-35.
An excellent introduction to the foundations of digital spectrum
analysis. These tutorials assume no mathematics beyond high
school algebra, trigonometry, and geometry. More advanced
mathematical tools (e.g., vector and complex number
manipulations) are developed as needed in these articles.
344
Marler, P. (1969). Tonal quality of bird sounds. In Hinde, R. A. (Ed.). Bird
vocalizations: Their relation to current problems in biology and
psychology (pp. 5-18). Cambridge University Press: Cambridge.
Includes an excellent qualitative discussion of how the time and
frequency analysis resolution of a spectrum analyzer interact with
signal characteristics to affect the “appearance” of a sound either
as a spectrogram or as an acoustic sensation.
Oppenheim, A.V. & Schafer, R.W. (1975). Digital signal processing. PrenticeHall: Englewood Cliffs, NJ.
A classic reference, written principally for engineers.
Rabiner, L.R. & Gold, B. (1975). Theory and Application of Digital Signal
Processing. Prentice-Hall, Englewood Cliffs, NJ.
Another classic engineering reference.
Yost, W.A. & Nielsen, D.W. (1985). Fundamentals of hearing: An introduction.
2d ed. Holt, Rinehart and Winston: New York.
A good general text on human hearing that includes some
discussion of the elementary physics of sound and an appendix
that introduces basic concepts of Fourier analysis.
345
346
Appendix C
Configuring Audio Input
About this appendix
This Appendix explains how to use controls provided by your operating
system to select and adjust the audio device from which Raven gets its
input when acquiring new signals.
Windows 98, 2000, and XP
Depending on which version of the Windows operating system you’re
using, the icons and windows you see may appear slightly different from
those shown here. However, the audio input controls function the same
way in all versions of Windows.
To access the Windows audio input and output controls, double-click on
the speaker icon that is displayed in the icon tray in the Windows task bar
(Figure C.1).
Figure C.1. Windows icon tray.
Figure C.1. The Windows icon tray. Double-click on the speaker icon
to display the Windows audio input and output controls. (You may see
different icons in the tray, depending on how your system is configured.)
The Volume Control dialog appears (Figure C.2), showing the volume and
balance settings for various audio output devices. To display the controls
for audio input devices, choose Options > Properties. In the Properties
dialog that appears (Figure C.3), the Mixer device drop-down menu lists
all of the sound cards installed in your computer, and lets you choose
which one to use for audio input. On many computers, only a single
sound card is installed; its name may differ from that shown in Figure C.3.
Click on Recording, check the boxes next to all of the sound input devices
that you may want to use on the list of volume controls at the bottom of
the dialog box, then click OK.
347
Appendix C: Configuring Audio Input
Figure C.2. Windows playback controls.
Figure C.2. The Windows Volume Control dialog as it’s initially
displayed, showing Playback controls. Controls for different devices
may be displayed, depending on system configuration.
Figure C.3. Audio control Properties dialog.
Figure C.3. The Properties dialog for the Windows audio playback and
recording devices. The Mixer device name on your computer may differ
from what’s shown here.
348
The Volume Control dialog will be replaced by the Recording Source
dialog (Figure C.4). To choose which recording source will be used, check
the Select box for that device. To adjust the recording level, move the
Volume slider control.
To record from a CD in Windows, you must select the recording source “Wave Out Mix” that appears in the audio playback Properties dialog (Figure C.3). After clicking OK, you
should click the Select checkbox corresponding to Wave Out
Mix in the Recording Control dialog.
Changes that you make in this dialog box are reflected immediately in the
signal reaching Raven. You can thus start a recorder running in Raven and
then use the Windows audio controls to adjust the recording level. See
Chapter 4, “Signal Acquisition (Recording)” for a discussion of how to set
the record level properly.
Figure C.4. Recording Source controls.
Figure C.4. The Recording Source controls. Controls for different
devices may be displayed, depending on system configuration.
Mac OS X
Launch the utility program called Audio MIDI Setup, located in /
Applications/Utilities. Audio MIDI Setup is supplied with Mac OS X.
Click on the Audio Devices tab to display controls for choosing a sound
input device and setting the input (recording) level (Figure C.5).
For Default Input, choose the sound input device that you plan to use with
Raven. In most cases, this will be Built-in Audio. Under Audio Input (left
349
side of window), choose the Source to use for input to the audio device.
In most cases, this will be Line In. The choices listed for Source correspond
to physical input jacks or ports on the computer or external audio device.
To adjust the recording level, move the Volume slider control.
Changes that you make to the recording level in Audio MIDI Setup are
reflected immediately in the signal reaching Raven. You can thus start a
recorder running in Raven and then use the Audio MIDI Setup controls to
adjust the recording level. See Chapter 4, “Signal Acquisition (Recording)”
for a discussion of how to set the record level properly.
Do not make changes to the audio input format (sample rate,
number of channels, and sample size) in Audio MIDI Setup
while Raven is running. Doing so may result in corrupted
recordings in Raven.
Figure C.5. Audio MIDI Setup
Figure C.5. The Audio Devices screen in the Audio MIDI Setup
program.
350
Appendix D
Detector Theory
About this appendix
This appendix introduces some theoretical concerns involving the use of
detectors including statistical tradeoffs associated with detection. This
information may be helpful in configuring the optimal settings for a
detector.
Tradeoffs involved with using detectors
Detectors allow a user to search for repetitions of a set of user-defined
signal characteristics such as a particular call or note produced by an
animal. Detectors have the advantage of being able to search and select for
a given signal within large datasets, perhaps months’ worth of data, in a
relatively short period of time compared to what it would take a human
operator to complete the same task. It should be noted, however, that
detectors are not necessarily able to capture 100% of the target signals and
this should be taken into account when deciding whether or how best to
use a detector. The accuracy of your detections may be diminished by any
or all of the following factors:
•High background noise
•Low signal strength
•High signal complexity
•Clutter
If your signal contains significant background noise, is not very loud, is
especially complex, or if there are other sources of noise that would be
relatively indistinguishable to the detector with a given set of parameters,
the detection accuracy could be compromised.
Results of detecting Two specific results could occur if your sound recordings are imperfect:
in less-than-perfect
recordings 1. You may miss some of the signals that were present in the recording (a
false negative, or FN).
2. You may have detections which are not part of the set of signals for
which you were searching (a false positive, or FP).
One way to quantify your error rate is to take a sub-sample of the data set
and manually browse for the signals you want. You can then run the
desired detector on the same data set. A comparison of the number of
selections from hand-browsing with the number of detections can then be
351
Appendix D: Configuring Audio Input
made. You will then have some idea of the accuracy that you will achieve
for the whole dataset.
In addition to comparing a hand-browsed dataset with a detector run, you
can also compare one detector against another or even compare different
parameter settings within the same detector. Comparing a sub-sample of
your dataset can therefore allow you to make a choice between using a
detector and hand-browsing, and will help you decide which detector,
with what parameters, is the best to use.
Receiver operating The Receiver Operating Characteristic (ROC) curve refers to the
characteristic percentage of true positives (TPs) captured and the percentage of FPs
captured.
Figure D.1 ROC curve
Figure D.1. ROC curve showing the difference between a ‘good’
detector and a ‘bad’ detector.
The ROC curve in Figure D.1 quantifies the trade-off between the
percentage of TP captured and the percentage of FP captured. It shows
that if we need to capture at least 70% of the TP, then using the ‘good’
detector, we would expect to also detect 20% of the FP. Using the ‘bad’
detector would select 70% of the FP.
The particular method you choose should be determined by your
objectives and the amount of time and effort you are prepared to put into
352
the task. Bear in mind that the most important thing to know is what type
of results you expect. Then, you can adjust the detector parameters
accordingly before beginning the detection process.
Changing a A good first step toward efficiently running a detector is to optimize the
detector parameter detector’s parameters with the interactive detection mode. The following
graph shows the effect of changing a detector parameter on the proportion
of desired signals captured (TP) and the number of missed signals (FN) as
well as the number of incorrect signals captured (FP).
Figure D.2 Changing a detector parameter
Figure D.2 (a) The signal-to-noise ratio is high enough that all of the TP
can be captured without capturing any of the true negatives (TN). (b)
The signal-to-noise ratio is lower so that it becomes impossible to
capture all the TP without either capturing some FP and/or rejecting
some of the FN. Changing the parameter setting alters the ratios of
TN:FP and FN:TP. (c) To capture all the TP, you will also capture most
of the FP. Because of the high number of FP that would have to be
“weeded out”, this is a time consuming task. (d) By not capturing any
TN, you will also be rejecting the majority of FN.
353
354
Index
INDEX
A
a (in file name templates), 161
About Raven..., 32
Acquiring a signal, 21
See Recording a signal
Acquisition
See also Recording a signal
Activate Selection (selection contextual menu), 143,
157
Activate Selection Mode, 35
Active selection, 143
Playback, 9, 211
Active view, 47–48
Active window, 77
Adaptive filter (Adaptive line enhancer), 148
in a Recorder, 102
Aggregate entropy measurement, 169
aif filename extension, 13
AIFF file format, 13, 87
Aliasing, 101, 320–322
Anti-aliasing filter, 322
Amplified recording, 102
Amplify, 26, 149
Amplify Sound dialog box, 149
Batch Amplify, 150
Multi channel sounds, 207
Amplitude detector, 283–287
Amplitude threshold parameter, 283
Configure, 256, 283
Delay compensation parameter, 283, 286
Smoothing parameter, 283
Analog-to-digital conversion, 319
Analysis frequency, 332
Analysis resolution, 335–337
Annotations, 179
Creating, 180
Deleting, 180
Entering values, 180–182
Fill Down, 181
Naming, 180
Renaming, 180
Saving in file name, 161
Audio file formats, 13
Audio Format Not Supported error, 107
Audio input
Configuring under Mac OS, 349
Configuring under Windows, 347–349
Audio input device
Choosing, 19
Plug-ins, 82, 320
Auto-apply checkbox
in Color Scheme Editor, 296
in Configure New Spectrogram dialog box, 117,
129
Auto-Paging, 191
Auto-scrolling (while making selections), 142
Average entropy measurement, 170
Average power measurement, 168
Averaging parameter, 127–128
Axis
Configuring, 52
Units, 49
Visibility and window presets, 73
B
Background color
Changing, 44, 304
Band Limited Energy Detector, 274–283
Bandwidth filter, 276
Bandwidth filter parameters, 276
Configure, 256, 274
Duration parameters, 275, 279
Example, 277
Exclusion band, 276
Exclusion band parameters, 276
Frequency parameters, 275, 278
Minimum occupancy parameter, 275
Minimum separation parameter, 275, 280
Noise block size parameter, 274–275, 281
Noise estimation, 274–275
Noise estimation parameters, 281
Noise hop size parameter, 276, 281
Noise percentile parameter, 274, 276, 281
SNR threshold parameter, 275, 281
Bandpass filter, 226
simulated by STFT, 332, 340–342
Bandwidth, 333
Bandwidth 90% measurement, 172
Bartlett window function, 114
Batch Amplify, 150
Batch channel export, 212
Batch correlation, 240–251
Correlation table, 242
Example, 246
Input files, 241
Output folder, 241
Saving output, 244
Spectrogram, 240
Batch detection, 254, 262
355
Index
Batch Filter, 150
Beamforming, 213
Beamogram parameters, 216
Beamogram slice view, 57, 214
Beamogram view, 56, 214
Beams, 215
Configuring, 215
Beams, 215
Bearing analysis, 213
Bearing angles, 215
Begin file measurement, 176
Begin Time measurement, 156, 166, 175
Beta (Kaiser window parameter), 117
Biased Normalization, 227
Bit depth, 323–324
Blackman window function, 114
BMP (graphics) file format, 18
Brightness (spectrogram), 68, 130, 132, 294
and Window presets, 73
Buffer (when recording), 81
Bug reports, 314–315
C
Canary
Opening files in Raven, 7
Cascade Windows, 78
Ceiling color, 130, 132
Ceiling value, 130, 132
Center frequency measurement, 171
Center Position button, 50–51
Center time measurement, 172
Channels
Batch export, 212
Hiding, 206
Rearranging, 208
Saving a subset, 209
Selecting, 208
Showing, 206
Visibility, 65
Channels list (side panel), 206
Check for updates, 32
Choose Measurements (View menu), 166
Clear
Active Selection, 145
All Selections (all tables), 145
All Selections (Table N), 145
Selection N, 145
Clip exporter, 104
Clip File Format, 298
Clip File Names, 298
Clipping (during recording), 96–97
Cloning selections, 166
Close
356
All files, 25
Selection table, 25
Sound file, 25
Collapse Selection Table button, 150
Color bar view, 57
Color choosers, 293
HSB, 293
RGB, 294
Swatches, 293
Color schemes, 69–71, 289–296
as Linkable properties, 65
Auto-apply checkbox, 296
Brightness, 294
Color choosers, 293–294
Editing, 71, 291–295
HSB color chooser, 293
Hue, 293
Opacity, 294
Presets, 295, 303
RGB Color chooser, 294
Saturation, 294
Spectrograms, 29
Swatches color chooser, 293
Colormap
Relative power, 71
See Spectrogram colormap
Components pane (side panel), 67
Components, window, 67
Configure Beamogram, 215
Configure New Beamogram dialog box, 216
Array positions, 217
Bearing coordinate system, 218
Bearing grid, 218
Bearings increase parameter, 218
Filtering, 217
Hop size parameter, 217
Maximum bearnig parameter, 218
Minimum bearing parameter, 218
Normalize channels, 218
Number of bearings parameter, 218
Record size parameter, 218
Reference bearing parameter, 218
Sound speed, 216
Stop band attenuation, 217
Time grid, 217
Transition region bandwidth, 217
View channels, 219
Configure New Recorder dialog box, 19
Advanced button, 99, 103
Buffer Size parameter, 84
Channels checkboxes, 83
Device drop-down list, 83
Directory parameter, 89
Raven 1.4 User’s Manual
Index
Display tab, 84–86
File Format tab, 86
File Name tab, 88
File Names tab, 88, 92
File Size parameter, 88
Input tab, 82–84
Latency parameter, 85
Rate Conversion tab, 101
recorder presets, 93
Retroactive Recording Offset parameter, 103–
104
Sample Format drop-down list, 84
Sample Rate drop-down list, 83
Sample Size parameter, 87
Schedule tab, 103
Simple button, 100, 103
Sound File(s) parameter, 89
Speed tab, 99
Start Time parameter, 91
Update Rate parameter, 85
Window Preset drop-down list, 86
Configure New Sound Window dialog box, 4, 74,
188, 195
Configure New Spectrogram dialog box, 115–117,
119–122, 129
3 dB bandwidth, 117
Apply button, 129
Auto-apply checkbox, 129
Averaging parameter, 127–128
DFT Size parameter, 122
Frequency (Grid Spacing) parameter, 122
Overlap parameter, 121
Preset menu, 129
Time parameter, 121
Window Size
Slider control, 117
Window Size parameter, 116–117
Window Type parameter, 113
Configure New Spectrogram Slice View dialog box,
58
Configure Paging dialog box, 191–192
Configure Recorder dialog box
See Configure New Recorder dialog box
Configure Recorder... (contextual menu), 98–99
Configure Selection Spectrum dialog box, 113
Configure Spectrogram dialog box, 113
See Configure New Spectrogram dialog box
Configure Spectrogram Slice dialog box, 113
Configure View Axes dialog box, 52
Contextual menus, 45, 67
Contrast (spectrogram), 68, 130, 132
Control points
Selection, 159
Copy (Edit menu), 12, 26, 146
Copy Image Of, 26
Copy Selected Rows (selection table contextual
menu), 162
Copy Selections, 162
Correlation, 31, 221
Batch correlation, 240–251
Batch correlation table, 242
Calculation, 222–223
Complex envelope, 238
Correlation view, 231
Demeaning spectrogram values, 233
Linear power values, 232
Logarithmic power values, 232
Playback, 231
Scrolling, 231
Spectrogram clipping, 235
Spectrogram correlation, 221, 229–237
Spectrogram parameters, 234
Types, 221
Waveform correlation, 221, 223, 237–240
Create a new view, 28
Create Selection Mode, 35, 142
Current Heap Size, 311
Current page
See under Paged sound windows
Cut (Edit menu), 26, 146
D
Date tags (in file name templates), 90
dd tag (in file name templates), 90
Deactivate Selection (View menu), 145
Decimation
Decimation Factor parameter, 101
During signal acquisition, 101
Usefulness of, 101
See also Sample rate conversion
Decimation Factor, 101
Default directories, 306
Default measurement precision, 307, 309
Default presets, 301–302
Delay compensation parameter (in Amplitude detector), 283, 286
Delete
Selection, 26, 146
View, 66
Delta Frequency measurement, 175
Delta power measurement, 168
Demeaning
in Spectrogram correlation, 233
Detection, 31, 253
Amplitude detector, 283–287
Clear selections, 260
357
Index
Configure, 256, 260, 274
Detector active, 260
Energy Detector, 274–283
Example, 267
Full, 254, 261
in a Recorder, 267
Interactive, 254–255
Naming, 256
Presets, 256, 273
Real-time, 254, 267
Run full, 260
Saving detector settings, 273
Saving selections, 273
Side panel, 257
Timing, 257
Detection tab (side panel), 40
Detector Active, 260
Device unavailable error (for signal acquisition),
107
DFT (discrete Fourier transform), 330–331
DFT size, 122–123, 332, 335, 339–340
Lock icon, 123
Parameter, 122–123
Digitizing error, 323
Directory parameter (recording to file(s)), 89
Discrete Fourier transform (DFT), 110, 330–331
Display tab (Configure New Recorder dialog box),
84, 86
Docking controls, 40
Documentation, 31
Duration 90% measurement, 174
Dynamic range, 323
E
Edit menu
Amplify..., 149
Copy, 12, 146
Copy Image Of, 18
Cut, 146
Delete, 146
Filter Active Selection With, 148
Filter Active Selection With > Other, 147
Filter All With, 148
Filter All With > Other, 147
Filter Around Active Selection, 147
Filter Around All, 147
Filter Out Active Selection, 147
Filter Out All, 147
Paste, 12, 146
Preferences, 27
Redo, 147
Select all, 26
358
Undo, 147
Edit toolbar, 33
Email (from within Raven), 315
End file measurement, 176
End Time measurement, 156, 166, 176
Energy Detector, 274–283
See Band Limited Energy Detector
Energy measurement, 169
EPS (graphics) file format, 18
Errors
Audio Format Not Supported, 107
Device unavailable, 107
Frame rate monitor, 107
Out-of-memory, 310
Expand Selection Table button, 150
Exporting clip files, 104
Exporting images, 25
Exporting samples, 25
Extender point, See under Selection control points,
142
F
False negative, 351
False positive, 351
Fast Fourier transform (FFT), 331
FFT (fast Fourier transform), 331
File boundary markers (in file sequence views), 197
File Format tab
Configure New Recorder dialog box, 86
Configure New Sound dialog box, 5
File formats
AIFF, 13, 87
For recording to file(s), 87
WAVE, 87
File menu
Close "Sound N", 25
Close all, 25
Close Selection Table, 25
Exit, 25
Export Image Of, 18
Export samples, 25
New Recorder..., 19, 82
New selection table, 152
New selection table (with measurements), 152
New Sound Window, 12
Open recent selection table, 156
Open Selection Table, 156
Open Sound File Sequence..., 195
Open Sound Files..., 2
Open Workspace..., 15
Print One Page of “Sound N”, 17
Print Pages of “Sound N”, 18
Printer Page Setup, 18
Index
Save “Sound N”, 14
Save “Sound N” As..., 14
Save Active Selection As..., 15, 160
Save All Selections..., 161
Save As List File, 15
Save Selected Channels As, 209
Save Selection Table As, 153
Save Selection Table As..., 155
Save Workspace As..., 15
File name collisions (recording to file(s)), 91
File Name tab (Configure New Recorder dialog
box), 88
File name templates, 89
File Names tab (Configure New Recorder dialog
box), 92
File naming (recording to file(s)), 88–90
File offset measurement, 176
File sequences, 194–197
Choosing files individually, 196
Individual file selection, 196
List files, 195–196
File toolbar, 32
Filter
Batch Filter, 150
Filterbank (model of STFT analysis), 332
Filtered RMS amplitude measurement, 175
Filtering, 147
Adaptive line enhancer, 102, 148
Around, 147
Out, 147
Selected frequency band, 147
Filters, 26
Applying a saved preset, 148
Defining your own, 147
Filter bandwidth, 117
Saving your own, 147
First quartile frequency measurement, 171
First quartile time measurement, 173
Floor color, 130, 132
Floor value, 130, 132
Format Column... (selection table contextual
menu), 177
FP, 351
Frame rate monitor, 107
Frames (in short-term Fourier transform), 332
Frequency (Grid Spacing) parameter, 122
Frequency 5% measurement, 172
Frequency 95% measurment, 172
Frequency analysis resolution, 333
Frequency domain, 328
Frequency grid spacing, 122–123, 339–340
Full detection, 254, 261
G
Generate Selections, 165
Grab And Scroll Mode, 35
Graphics export
See Image export, 18
Gray background (in views), 49
Grid lines, 50
Display, 50
in Sound window presets, 73
Major, 50
Minor, 50
Visibilty, 73
Grid spacing, 335
See also Time grid spacing, Frequency grid
spacing
Group By Time, 76–77
Group By View, 75–76
Grouping (lines, views, channels), 75–77
Grouping options (side panel, Layout tab), 75–77
H
Hamming window function, 114
Hann window function, 114
Hanning window function, 114
Heap, 310
Help menu
Email Feedback, 315
Raven Pro FAQ, 314
Raven Pro help forum, 314
hh tag (in file name templates), 90
Hidden, Side panel, 40
Hide all button, 41
Hide All Channels, 206
Hide All Views, 42
Hide View (View menu), 66
Hiding/showing channels, 206
High Frequency measurement, 156, 166, 176
Highlighting selections, 161
Hop Size parameter, 120–122
HSB color chooser, 293
Hue, 293
I
Image export
to file, 18
via clipboard, 18
Information tab (side panel), 40, 42
Interactive detection, 254–255
Clear selections, 260
Configure, 255–256
Detector active, 260
359
Index
Multiple detectors, 260
Naming, 256
Paged sounds, 259
Presets, 273
Saving detector settings, 273
Saving selections, 273
Side panel, 257
Timing, 257
IQR (Inter quartile range) bandwidth measurment,
172
IQR (Inter quartile range) duration maeasurement,
173
Hop Size (in Configure Spectrogram dialog
box), 122
Lock icons (in Configure New Spectrogram dialog box), 122
Overlap (in Configure Spectrogram dialog
box), 122
Toolbar, 32
Look and Feel
Changing, 44, 304
Looping playback, 10
Low Frequency measurement, 156, 166, 176
J
Magnitude spectrum, 329–330
Max amplitude measurement, 174
Max bearing measurement, 177
Max Frequency measurement, 170
Max Power measurement, 170
Max time measurement, 175
Maximum heap size, 310
Configuring, 311
Resetting outside Raven, 311
See also Memory allocation
Measurement Chooser dialog box, 166–167, 178
Measurement list
Measurements, 29, 151, 166
Choosing, 166
Default precision, 307, 309
Default presets, 302
Format, 177
Plotting, 167
Precision, 177, 307, 309
Presets, 177–178, 302
Visibility, 167
Memory allocation, 310
Heap size, 310
Maximum heap size, 310
Out-of-memory errors, 310
Memory Manager, 309–313
Current Heap Size, 311
Memory Available, 311
Memory usage list, 312
Memory usage summary, 311
Memory Used, 311
Reclaim button, 313
Reclaiming unused memory, 313
Memory usage list, 312
Memory usage summary, 311
Memory Used, 311
Menu bar, 1–2, 23
Min amplitude measurement, 174
Min time measurement, 175
Java Virtual Machine (JVM), 310
JPEG (graphics) file format, 18
JVM (Java Virtual Machine), 310
K
Kaiser window function, 114, 117
Keyboard shortcuts, 38
L
Lag, 221
Landscape orientation (for printing), 18
Layout tab (side panel), 40, 42, 47
Channels list, 206
Components list, 67
Grouping options, 75
Hiding/showing window components, 67
Lines box, 75
Length measurement, 176
Line enhancement
in a Recorder, 102
Line titles, 67–68
Lines box (side panel, Layout tab), 75
Linkable properties, 64–65
Linkage (of views), 50, 62, 64–66
Creating, 64
Linkable properties, 64–65
Linkage groups, 64–65
Linked spectrographic views, 65
Restoring, 64
Unlinking views, 63
Linkage tab (side panel), 64, 77
List files (for file sequences), 92, 195–196
ll tag (in file name templates), 90
Lock
DFT Size, 123
360
M
Index
mm tag (in file name templates), 90
Mouse measurement field, 43–44
Move View Down (View menu), 67
Move View Up (View menu), 67
Moving selections, 162
mp3, Cannot play variable bitrate files, 6
Multi-channel sounds, 207, 210
Editing, 207
Hiding/showing channels, 205–206
Playback, 210
Selecting channels, 210
Multiple-line views, 74–75
grouping, 75
N
n (in file name templates), 90
New
Beamogram slice view, 57
Beamogram view, 57
Recorder, 24, 82
Selection spectrum view, 112
Selection Table, 152
Selection table (with measurements), 152
Sound Window, 12, 24
Spectrogram slice view, 112
Spectrogram view, 112
View, 57
Normalization, 226
Biased, 227
Subset, 228
Unbiased, 227
Number of Lines
As linkable property, 65
Changing, 75
Nyquist frequency, 320–322
O
Online Resources, 314
Opacity, 294
Open
.mp3, .mp4, .aifc, .mov formats, 6
Canary files, 7
CD audio tracks, 6
Multiple sound files, 198
Recent Folder, 6
Recent selection table, 156
Recent selection table folder, 25
Recent Sound File, 6
Selection Table, 156
Sound File Sequence, 195
Sound Files, 2–3
Workspace, 25
Open Sound File Sequence dialog box, 195
Open Sound Files dialog box, 2–3, 198–199
Oscillogram, 7
Overlap, 335
Overlap parameter, 120–121
in Correlation, 234
Overwriting files (during acquisition), 309
P
Padlock icons (in Configure New Spectrogram dialog box), 122
Page Back button, 190
Page Forward button, 190
Page increment, 188, 190
Reconfiguring, 191
Page navigation panel
See under Paged sound windows
Page setup, 25
Page size, 188
Reconfiguring, 191
Page, See Paged sound windows
Paged sound windows, 187
Auto-paging, 191
Configuring, 188
Current file indicator, 190
Current page, 187
Editing prohibited, 187–188
Jump to file, 190
Jump to time, 190
Page Back button, 190
Page Forward button, 190
Page increment, 188, 190–191
Page navigation panel, 189–190
Page size, 188, 191
Paging scrollbar, 190
Recent pages, 191
Selectins in, 191
Step Back button, 190
Step Forward button, 190
Step increment, 188, 190–191
Paging scrollbar, See underPaged sound windows
Paste (Edit menu), 12, 26, 146
PCM (Pulse Code Modulation), 84
Peak amplitude measurement, 174
Peak correlation measurement, 177
Peak frequency measurement, 170
Peak lag measurement, 177
Peak power measurement, 170
Peak time measurement, 175
Performance tips, 314
Phase, 329
Phase spectrum, 329–330
Pitch tracking, 183
361
Index
Play
Looping playback, 10
Multi-channel sounds, 210
Recording, 96
Reverse playback, 10
Selection, 9, 211
Selection spectrum, 9
Sound file, 36
Sounds in correlation, 231
Spectrogram slice, 9
Visible, 9
Play button, 8
Play toolbar, 36, 211
Playback cursor, 9
Configuring, 9
Playback rate, 11, 37
Playback tab (side panel), 40, 210
Playing a sound, 8
Playing while recording, 96
PNG (graphics) file format, 18
Point selections, 141, 144
Portrait orientation (for printing), 18
Position (of a view), 51
Setting, 52
vs. Position location, 51
Position location, 51
vs. Position, 51
Position markers, 10, 50, 52
Power Spectra Calculation
Comparing two methods, 305
New method, 112, 304
Power to color correspondance in colormaps, 71
Preference file, 300
Preferences, 299–317
Default directories, 306
Default measurement precision, 307, 309
Default presets, 301–302
Overwriting files (during acquisition), 309
Power spectra calculation method, 299
Preset Manager, 297
Preset menu
See Presets, 129
Presets
Audio File Format, 298
Audio File Names, 298
Audio Recorder, 298
Audio Recorder Display, 298
Audio Recorder Input, 298
Audio Recorder Schedule, 298
Color Scheme, 298
Color scheme, 295
Detector, 298
Filter, 298
362
Measurement, 177–178
Measurement List, 298
Recorder, 93
Sample Rate Conversion, 298
Selection Labels, 298
Sound Window, 298
Spectrogram, 129
Spectrogram Parameters, 298
Speed Correction, 298
Types, 298
Window, 73
Window, precedence over other presets, 74
Print orientation, 18
Printing, 17
Sound file, 25
Progress bar, 95–96
Pulse Code Modulation (PCM), 84
Q
Quantization noise, 323
QuickTime
On Mac OS, 6
On Windows, 6
Opening sound files, 6
QuickTime installer, 32
Downloading, 6
R
Range selections, 141–142
Rate Conversion tab (Configure New Recorder dialog box), 101
Raven, 314
Raven desktop, 2
Raven license agreement, 32
Raven Pro help forum, 32
Raven User’s Manual, 31
Raven web site, 31
Raven window, 1
RavenPreferences.txt file
See Preference file
Read about updates, 32
Real-time detection, 254, 267
Example, 267
Real-time views (during signal acquisition), 21, 95
Waveforms, 21
Reclaim button, 313
Reclaiming unused memory, 313
Record button, 19, 82
Record length, 333
Record To drop-down list, 82
Recorder presets, 93
Default, 302
Index
Recorder windows, 21, 81, 94–95, 98, 104
Editing in, 98
Progress bar, 95–96
Reconfiguring, 98
Record-to-Disk button, 94, 96
Record-to-Memory button, 94–95
Stop Recording button, 95
Stop-Recording-to-Disk button, 96
Recorders
Creating, 82
Parameters, 82
Recording a signal, 18–19, 21, 81
Amplifying, 102
Buffer size, 81, 84
Channels, choosing, 83
Display configuration, 84–86
Display latency, 85
Exporting clip files, 104
File format, 87
File naming, 88
File size, 88
Input configuration, 82–84
Latency, 85
Processing latency, 85
Recorder presets, 93
Recording buffer, 85
Recording level, 96
Recording parameters, 82
Retroactive, 104
Sample format, 84
Sample rate, 83
Sample rate conversion during, 101–102
Sample size, 84, 87
Scheduled recording, 103
Speed correction, 99
to a File sequence, 92
to File(s), 86, 96
To memory, 19
to Memory, 95
Update rate, 85
Window presets and recorder windows, 86
See also Configure New Recorder dialog box;
Recording modes
Recording modes, 81–82
Choosing, 82
Record to file, 82
Record to file sequence, 82
Record to memory, 81
Records
In short-time Fourier transform, 332
in Spectrographic views, 110
Record-to-Disk button, 94, 96
Record-to-File button, 104
Record-to-Memory button, 21, 94–95
Rectangular window function, 114
Redo (Edit menu), 26, 147
Rendering data
Entries in Memory Manager, 312
Resize the side panel, 41
Retroactive recording
Offset parameter, 103–104
To files, 104
RGB Color chooser, 294
RMS amplitude measurement, 174
Run Full, 260
S
Sample frame, 88
Sample rate, 320–321, 323
Sample rate conversion
During recording, 101–102
Sample size, 84, 323–324
Samples
Internal floating-point representation, 324
Sampling, 319
Saturation (in HSB color chooser), 294
Save
Active Selection As..., 160
All Selections, 161
Channels, 209
Selection table, 153
Save Color Scheme Preset dialog box, 296
Save Sound Window Preset dialog box, 73
Scale (of a view), 52, 54
Default, 52
resetting, 55
Setting, 52
Zoom in, 54
Zoom out, 54
Scheduled recording, 103
Scroll thumb, 11, 48
Scrollbars, 48
Paging, 190
Scrolling playback, 9–10
Select All, 26
Selectable look and feel, 44, 304
Selection, 150
Selection bounds, 156–157, 166
Selection control points, 142
Active, 142
Extender point, 142
Selection files, 156
Format, 156
Retrieving, 156
363
Index
Selection ID, 142, 150
Selection labels, 142
Configuration, 73
Visibility, 73
Selection modes, 35
Activate Selection Mode, 35, 143
Create Selection Mode, 35, 142
Grab And Scroll Mode, 35
Selection number, 150
Selection spectrum views, 8, 56, 109–110, 112, 139
Creating with linked parameters, 65
Play, 9
Selection tab (Side panel), 159
Selection tables, 150, 155, 166
Activate Next Selection button, 157
Activate Previous Selection button, 157
and Paged sound windows, 160
Closing, 153
Collapsing, 150
Configuring, 154–155, 178
Copying rows, 161–162
Creating, 152
Down arrow button, 157
Expanding, 150
Saving, 153, 155
Sort order, 154
Splitting, 157
Up arrow button, 157
Visibility of entries, 153
Selections, 9, 141–142, 145–146, 155, 166
Activating, 143, 157
Active, 143
Bounds, 166
Clearing, 145
Cloning, 166
Committed, 143
Copying, 12, 146, 162
Creating, 141–142
Cutting, 146
Deactivating, 145
Defining, 141–142
Deleting (data), 146
Drawing, 153
Finding, 158
Highlighting, 161
ID, 150
Inactive, 143
Modifying, 142
Modifying bounds, 158
Modifying channels, 160
Moving, 162
Multi-channel, 213
Number, 150
364
Pasting, 12, 146
Point, 141, 144
Range, 141–142
Removing, 145
Renumbering, 146
Retrieving, 156
Saving active selection, 160
Saving all, 161
Spanning page boundaries, 191
Uncommitted, 143
Viewing, 158
Visibility, 153
Sequence tags (in file name templates), 90
Short-time Fourier transform (STFT), 332
Show all button, 41
Show All Channels, 206
Show All Views, 42
Showing/hiding Channels, 206
Side panel, 1–2
Detection tab, 40
Docking, 40
Hiding, 40
Information tab, 42
Layout tab, 67
Linkage tab, 64
Playback tab, 40, 210
Resizing, 41
Selection tab, 159
Sidelobe rejection, 114, 343
Sidelobes, 113, 340–342
Signal acquisition, 81
See Recording a signal
Smooth Spectrogram (View menu), 135–137
SMTP (Standard Mail Transfer Protocol) server, 316
Sonagrams, 332
Sound files
AIFF format, 13
Opening, 2–3
Opening multiple, 198–199
Opening recent, 6
WAVE format, 13
Sound pressure, 327–328
Sound windows
Activating, 77
Cascade, 78
Closing, 47
Default presets, 303
Minimizing, 47
Moving, 47
Paged, 187
Printing, 17
Properties, applying to all, 80
Resizing, 47
Selecting, 30
Index
Tile, 78
Tile Horizontally, 78
Tile vertically, 78
Spectral smearing, 340–342
Spectrogram caching
Memory Used, and, 312
Spectrogram clipping
in Correlation, 235
Spectrogram colormap, 70
Custom, 296
Spectrogram correlation, 221, 229–237
Batch, 240
Calculation, 222
Demeaning, 233
Linear, 232
Logarithmic, 232
Relevance, 225
Spectrogram clipping, 235
Spectrogram parameters, 234
Spectrogram Parameters... (View menu), 140
Spectrogram presets, 129
Default, 301
Spectrogram slice views, 8, 44, 56, 66, 109–110, 137–
138
Blank, 59
Linkage to spectrogram views, 66
Play, 9
Significance of spectrum values, 138
Time position, 137
Time position of, 66
“Hidden” time axis, 66
Spectrogram smoothing, 334–335
Spectrogram toolbar, 37
Spectrogram views, 7, 56, 109–110, 129–130, 132,
134–137
Brightness, 68, 130, 132
Caching of data, 312
Changing parameters, 140
Colormap, 70
Contrast, 68, 130, 132
Creating, 68
Linkage to spectrogram slice views, 66
Parameters
Significance of color (grayscale) values, 129
Smoothing, 134–137
Speed of calculation, 325
Time alignment of data, 132–134
Spectrograms, 133, 135–136, 332, 334–336, 338–343
Spectrum averaging, 127–128
Spectrum source data, 132
Spectrum time, 132
Speed correction, 5
Recording, 99
Speed tab (Configure New Recorder dialog box), 99
split a selection, 163
ss tag (in file name templates), 90
Start time (for naming recorded files), 91
Step Back button, 190
Step Forward button, 190
Step increment, 188, 190
Reconfiguring, 191
STFT (short-time Fourier transform), 332
Stop [playback] button, 9
Stop Recording button, 21, 95
Stop-Recording-to-Disk button, 96
Storage requirements, 325
Support (for Raven), 315
Swatches color chooser, 293
T
Tags (in file name templates), 90
Technical support, 315
Third quartile frequency measurement, 171
Third quartile time measurement, 173
3 dB bandwidth
See Filter bandwidth
TIFF (graphics) file format, 18
Tile Windows, 78
Time 5% measurement, 173
Time 95% measurement, 174
Time analysis resolution, 333
Time axis marker, 10
Time axis position marker, 10
Time domain, 328
Time grid spacing, 121, 335, 338–339
Time grid spacing (hop size), 120
Time offset, 221
Time parameter, 121
Time tags (in file name templates), 90
Time-frequency uncertainty principle, 333–337
Toolbars, 1–2
Edit toolbar, 33
File toolbar, 32
Lock/unlock, 32
Play toolbar, 36, 211
Spectrogram toolbar, 37
View toolbar, 34
Tools menu
Amplitude detector, 254
Band Limited Energy Detector, 254
Batch Adaptive Filter, 150
Batch Amplify, 150
Batch Band Filter, 150
Batch channel export, 212
365
Index
Batch correlator, 240
Correlator, 225
Detector, 254
Tooltips, 1
TP, 352
Triangular window function, 114
Troubleshooting, 314
True positive, 352
U
Unbiased Normalization, 227
Uncertainty principle
See time-frequency uncertainty principle
Undo (Edit menu), 26, 147
Undocked, 40
Unlink View (View menu), 63
Unlinking views, 62–63
Unlock
Toolbar, 32
Unrecoverable errors, 316
Updates
Finding, 32
Software and documentation, 313
V
Variable bitrate mp3 files, 6
View menu
Active selection > Find, 158
Apply To All, 80
Choose Measurements, 166
Clear Active Selection, 145
Clear All Selections, 145
Color Scheme > Colormap-name, 70
Color Scheme > Edit..., 71, 292
Color Scheme > Reverse Color Map, 70–71
Configure grids, 50
Configure selection labels, 182
Configure selection visibility, 153
Configure view axes, 52
Deactivate Selection, 145
Delete View, 66
Find selection, 158
Hide View, 66
Highlight rows, 161
Interactive detection, 254
Move View Down, 67
Move View Up, 67
New > Beamogram slice view, 57
New > Similar Selection Spectrum View, 65
New > Similar Spectrogram Slice View, 65
New > Similar Spectrogram View, 65
New > Spectrogram Slice View..., 57
366
New view, 57
Smooth Spectrogram, 135–137
Spectrogram Parameters..., 140
Unlink View, 63
Window Preset > Preset-name, 73
Window Preset > Save “Preset-name”, 74
Window Preset > Save As..., 73
View selection button, 47
View toolbar, 34
Views
Creating, 57
Deleting, 66
Display order of, 67
Multiple lines, 74–75
W
wav filename extension, 13
WAVE file format, 13, 87
Waveform correlation, 221, 223, 237–240
Calculation, 223
Complex envelope, 238
Waveform view, 7, 56
What’s New, 31
Window
See Sound windows
Window components
Hiding, 67
See Components, window
Showing, 67
Window functions, 113, 116–117, 342–343
Bartlett, 114
Blackman, 114
Hamming, 114
Hann, 114
Hanning, 114
Rectangular, 114
Triangular, 114
Window menu
Background Color..., 44, 304
Cascade windows, 78
Look and Feel, 44, 304
Memory Manager, 309
Preset manager, 298
Tile windows, 78
Tile windows horizontally, 78
Tile windows vertically, 78
Tooltip color, 44
Zoom window, 30, 60, 158
Window overlap, 121, 338–339
Window presets, 4, 28, 73
and Recorder windows, 86
Precedence over other presets, 74
Index
Window size, 73
Choosing, 337
Window Size parameter, 116–117
in Correlation, 234
Window Size slider control, 117
Window Type parameter, 113
Windows
Bring to front, 30
Windows menu, 78
Workspace files
Opening with missing/moved sound file(s), 16
Workspace Resolver Dialog, 17
Workspaces, 15
Default, 305
Opening, 15
Opening (drag and drop), 15
Saving, 15
wsp filename extension, 15
Y
yy tag (in file name templates), 90
Z
Zoom In button, 54
Zoom Out button, 54
Zoom window, 30, 60, 158
Zoom-to-All button, 55
Zoom-to-Selection button, 55
367

Raven Pro 1.4 User`s Manual - Cornell Lab of Ornithology

Transcription

Similar documents

Lucy Raven - School of Art

Will It Fly? | Decorati Access

May 2015 - Raven Home School

December 2010 Issue

Lucy Raven - Hammer Museum

AccuBoom™ Case IH SPX 3310 CAN Installation

61st Mason Dixon - Loch Raven Skeet and Trap Center

YKSD Connections - Yukon-Koyukuk School District

May 2012 - Mount St. Mary`s University

Raven Issue No. 17